ORIGINAL PAPER

A Phylogeny and Timescale for Marsupial Evolution Basedon Sequences for Five Nuclear Genes

Robert W. Meredith & Michael Westerman & Judd A. Case &

Mark S. Springer

# Springer Science + Business Media, LLC 2007

Abstract Even though marsupials are taxonomically less diverse than placentals, they exhibitcomparable morphological and ecological diversity. However, much of their fossil record isthought to be missing, particularly for the Australasian groups. The more than 330 livingspecies of marsupials are grouped into three American (Didelphimorphia, Microbiotheria, andPaucituberculata) and four Australasian (Dasyuromorphia, Diprotodontia, Notoryctemorphia,and Peramelemorphia) orders. Interordinal relationships have been investigated using a widerange of methods that have often yielded contradictory results. Much of the controversy hasfocused on the placement of Dromiciops gliroides (Microbiotheria). Studies either support asister-taxon relationship to a monophyletic Australasian clade or a nested position within theAustralasian radiation. Familial relationships within the Diprotodontia have also proveddifficult to resolve. Here, we examine higher-level marsupial relationships using a nuclearmultigene molecular data set representing all living orders. Protein-coding portions of ApoB,BRCA1, IRBP, Rag1, and vWF were analyzed using maximum parsimony, maximumlikelihood, and Bayesian methods. Two different Bayesian relaxed molecular clock methodswere employed to construct a timescale for marsupial evolution and estimate the unrepresentedbasal branch length (UBBL). Maximum likelihood and Bayesian results suggest that the root ofthe marsupial tree is between Didelphimorphia and all other marsupials. All methods providestrong support for the monophyly of Australidelphia. Within Australidelphia, Dromiciops is thesister-taxon to a monophyletic Australasian clade. Within the Australasian clade, Diprotodontiais the sister taxon to a Notoryctemorphia + Dasyuromorphia + Peramelemorphia clade. Within

J Mammal EvolDOI 10.1007/s10914-007-9062-6

The online version of this article (doi:10.1007/s10914-007-9062-6) contains supplementary material, which isavailable to authorized users.

R. W. Meredith (*) :M. S. Springer (*)Department of Biology, University of California, Riverside, CA 92521, USAe-mail: rmere001@student.ucr.eduurl:

M. WestermanDepartment of Genetics, La Trobe University, Bundoora, Victoria 3083, Australia

J. A. CaseCollege of Science Health and Engineering, Eastern Washington University, Cheney, WA 99004, USA

the Diprotodontia, Vombatiformes (wombat + koala) is the sister taxon to a paraphyleticpossum group (Phalangeriformes) with kangaroos nested inside. Molecular dating analysessuggest Late Cretaceous/Paleocene dates for all interordinal divergences. All intraordinaldivergences were placed in the mid to late Cenozoic except for the deepest splits within theDiprotodontia. Our UBBL estimates of the marsupial fossil record indicate that the SouthAmerican record is approximately as complete as the Australasian record.

Keywords Marsupialia . Phylogeny . Fossil record . Molecular divergence dates .

Ameridelphia . Australidelphia . Unrepresented basal branch length

Introduction

Marsupialia and Placentalia comprise the two major groups of living mammals. Marsupials aretaxonomically less diverse than placentals by an order of magnitude, but their long, and oftenisolated, evolutionary history has resulted in an assemblage of species whose morphological andecological diversity is nearly comparable to that seen in placental mammals (Springer et al. 1997a).Notable exceptions include the absence of marsupial analogs to Cetacea and Chiroptera.

There are more than 330 species of recent marsupials (Wilson and Reeder 2005). Wilsonand Reeder (2005) divide these species into three American orders [Didelphimorphia (17genera, 87 species), Microbiotheria (1 genus, 1 species), Paucituberculata (3 genera, 6 species)]and four Australian orders [Dasyuromorphia (23 genera, 71 species), Diprotodontia (39 genera,143 species), Notoryctemorphia (1 genus, 2 species), Peramelemorphia (8 genera, 21 species)].Szalay (1982) proposed a division of these orders into the cohorts Australidelphia andAmeridelphia based primarily on the distinction between the continuous lower ankle jointpattern (CLAJP) and the separate lower ankle joint pattern (SLAJP). CLAJP is the derivedcondition and characterizes Australidelphia, which is comprised of the four Australasian ordersplus the American order Microbiotheria. SLAJP is the primitive condition and characterizesAmeridelphia, which includes the American orders Didelphimorphia and Paucituberculata.Australidelphia has subsequently been corroborated by analyses of morphological (Luckett1994; Szalay and Sargis 2001; Horovitz and Sánchez-Villagra 2003), molecular (Kirsch et al.1991, 1997; Springer et al. 1998; Phillips et al. 2001, 2006; Amrine-Madsen et al. 2003), andmixed data sets (Asher et al. 2004). Molecular data sets confirming Australidelphia are diverseand include single-copy DNA hybridization (Kirsch et al. 1991, 1997), mitochondrial genomesequences (Phillips et al. 2001; Nilsson et al. 2003, 2004; Munemasa et al. 2006),concatenations of nuclear gene segments (Amrine-Madsen et al. 2003), and mixedmitochondrial–nuclear data sets (Phillips et al. 2006).

Resolving relationships within Australidelphia has proved difficult and the above-mentioned data sets offer contradictory results. For example, some analyses nest micro-biotheres within the Australasian radiation (Kirsch et al. 1997; Burk et al. 1999; Szalay andSargis 2001; Nilsson et al. 2003, 2004), whereas in other studies microbiotheres are the sister-taxon to a monophyletic Australasian clade (Amrine-Madsen et al. 2003; Phillips et al. 2006).Resolving the relationship of microbiotheres relative to other australidelphians is critical forunderstanding the early biogeographic history of Australidelphia (Kirsch et al. 1991; Springeret al. 1998; Szalay and Sargis 2001; Nilsson et al. 2004).

Subsequent to the proposal of Ameridelphia by Szalay (1982), Temple-Smith (1987)suggested that this cohort might be a monophyletic sister-group to Australidelphia based on theoccurrence of epididymal sperm-pairing in the former. Alternatively, analyses of mitochondrialgenomes (Nilsson et al. 2003, 2004) and concatenated nuclear gene sequences (Amrine-

J Mammal Evol

Madsen et al. 2003) suggest that Ameridelphia is paraphyletic, with a basal split betweenDidelphimorphia versus Paucituberculata + Australidelphia. Statistical tests reported byAmrine-Madsen et al. (2003) could not discriminate between rooting Marsupialia betweenDidelphimorphia and other marsupials and between Australidelphia and Ameridelphia.However, Asher et al. (2004) found significant statistical support for rejecting a root betweenAustralidelphia and Ameridelphia. Resolving the root of the marsupial tree remains critical forinferring the geographic provenance of the last common ancestor of Marsupialia.

The construction of a molecular timescale for marsupial evolution requires the integration offossil dates and molecular sequence data. Studies employing fossil-calibrated molecular clockswere previously the standard for estimating molecular divergence times. The divergencebetween marsupials and placentals is minimally 125 Ma based on the oldest undisputedmetatherian fossil (Sinodelphys; Luo et al. 2003). Kumar and Hedges (1998) estimated thatmarsupials and placentals diverged 173 Mya using a molecular clock for multiple nucleargenes that was calibrated against the split between birds and mammals at 310 Ma. Penny et al.(1999) obtained a similar estimate for this split (176 Ma) using two calibrations (horse to rhinoat 55 Ma; birds to mammals at 310 Ma) and linear interpolation along the backbone of themammalian tree. The last common ancestor of crown-group metatherians (i.e., Marsupialia) isostensibly much younger. Kirsch et al. (1997) obtained a date of 72 Ma for the last commonancestor of marsupials based on rate-calibrated DNA–DNA hybridization data. Springer (1997)used rate-adjusted 12S rRNA distances and concluded that cladogenesis between marsupialorders was mostly centered on the Cretaceous/Tertiary (K/T) boundary at approximately65 Ma. Recent fossil discoveries suggest that crown-group Metatheria has a minimum age ofLancian (∼69–65 Ma) or possibly Judithian (∼79–73 Ma) depending on the phylogeneticposition of polydolopimorphs (Case et al. 2005; Goin et al. 2006).

More recently, parametric (e.g., Thorne et al. 1998; Kishino et al. 2001) and semi-parametric (e.g., Sanderson 2002) divergence dating methods that relax the molecular clockassumption have been employed to examine marsupial divergences (Hasegawa et al. 2003;Nilsson et al. 2003, 2004; Woodburne et al. 2003). These methods often perform better thanmethods that assume a strict molecular clock (Yang and Rannala 2006; Smith et al. 2006;Benton and Donoghue 2007). Reliable calibration points are essential for obtaining accurateestimates of divergence times with relaxed clock methods and this topic has receivedconsiderable attention (Yang and Rannala 2006; Benton and Donoghue 2007). Minimumconstraints on divergence times require (1) a fossil with diagnostic apomorphies for a particulargroup, (2) an accurate phylogenetic tree, and (3) a minimum geologic date for the fossil-bearing stratum. If all of these conditions are met, then the probability of an earlier divergencetime drops immediately to zero (Benton and Donoghue 2007). Benton and Donoghue (2007)advocate a “hard” bound for minimum divergence age constraints when these conditions aresatisfied. Alternatively, a “soft” bound for minimum ages may be more appropriate if there arepotential problems with fossil identification, tree robustness, and geologic dates for the fossil-bearing stratum. Maximum constraints on divergence times are more difficult to specify(Benton and Ayala 2003; Hedges and Kumar 2004; Reisz and Müller 2004; Benton andDonoghue 2007). Reisz and Müller (2004) and Müller and Reisz (2005) suggest the use ofphylogenetic bracketing to constrain maximum fossil divergence dates. Benton and Donoghue(2007, p. 28) advocate soft bounds for maximum divergence ages and suggest a pluralisticapproach that combines phylogenetic bracketing with stratigraphic bounding (e.g., “consider-ation of the absence of fossils from underlying deposits”). These authors urge that bothminimum and maximum constraints be fully substantiated so that as new fossils are foundand geologic dates are refined, new analyses can reflect the new information. Yang andRannala (2006) have shown that soft and hard bounds yield similar results when fossil

J Mammal Evol

calibrations are consistent with each other and with molecular data. However, soft-boundedconstraints perform better than hard-bounded constraints when poor fossil calibrations areused. This is because soft-bounded constraints allow sequence data to correct poor fossilcalibrations, whereas hard-bounded constraints are fixed and are impossible to overcome withany amount of sequence data.

Woodburne et al. (2003) used the estbranches and divtime5b programs of Thorne et al.(1998) and Kishino et al. (2001) with hard constraints and amino acid sequences for twoproteins (BRCA1 and IGF2) and obtained point estimates of 182–190 Ma for the split betweenmarsupials and placentals. Nilsson et al. (2003, 2004) used Sanderson’s (2002) penalizedlikelihood approach to estimate divergence times among marsupial orders and obtained a datefor the base of Marsupialia slightly after the K/T boundary at 64 Mya (Nilsson et al. 2003) orjust prior to the K/T boundary at 69 Ma (Nilsson et al. 2004). Nilsson et al. (2003, 2004) usedthree fixed calibration points, including 135 Ma for the split between marsupials andplacentals. They also used a tree on which marsupials and monotremes are sister taxa.Hasegawa et al. (2003) employed the relaxed molecular clock method of Thorne and Kishino(2002) with mitochondrial genome data and hard constraints and obtained a date ofapproximately 100 Ma for the split between didelphimorphians and australidelphians.Hasegawa et al. (2003) cautioned that this seemingly too old split may have resulted fromsparse marsupial sampling in their study and/or a model for changes in evolutionary rates thatdid not well describe marsupial history.

In the present paper we examine higher-level marsupial relationships using a molecular data setthat builds on the concatenation of multiple nuclear gene segments presented by Amrine-Madsenet al. (2003) and present a timescale for marsupial evolution based on the relaxed molecularclock approaches of Thorne and Kishino (2002) and Drummond et al. (2006) with these nucleargene sequences. These estimated dates of divergence are then used to estimate the unrepresentedbasal branch length (UBBL) of the marsupial fossil record following Teeling et al. (2005).

Materials and methods

Taxon sampling

Our study included 22 marsupials and nine placental outgroups, all of which are indicated inTable 1. Four of 22 marsupial taxa were chimeric above the genus level (Table 1). We followthe classification of Wilson and Reeder (2005) for marsupial orders and families, with theexception that we recognize two families (Didelphidae and Caluromyidae) within the OrderDidelphimorphia (Kirsch and Palma 1995). Marsupials included in our study represent allextant orders (sensu Wilson and Reeder 2005); placental taxon sampling includedrepresentatives of the four major clades (i.e., Afrotheria, Euarchontoglires, Laurasiatheria,and Xenarthra; Murphy et al. 2001) (Table 1).

Gene sequences

DNA was extracted using DNeasy Tissue extraction kits from QUIAGEN or using themethodology of Kirsch et al. (1990). Portions of exon 26 of ApoB (Apolipoprotein B), exon11 of BRCA1 (breast and ovarian cancer susceptibility gene 1), exon 1 of IRBP(interphotoreceptor retinoid binding protein gene), intronless Rag1 (recombination activatinggene-1), and exon 28 of vWF (von Willebrand factor gene) were amplified as describedelsewhere (Amrine-Madsen et al. 2003). External forward and reverse primers new to this study

J Mammal Evol

Table 1 Ordinal representation of genera included in this study

Marsupialia (Infraclass Metatheria)

Order Didelphimorphia

Didelphidae

Didelphinae (Didelphis/Lutreolina)

Marmosinae (Monodelphis)

Caluromyidae (Caluromys)

Order Paucituberculata

Caenolestidae (Caenolestes, Rhyncholestes)

Order Microbiotheria

Microbiotheriidae (Dromiciops)

Order Dasyuromorphia

Dasyuridae

Sminthopsinae (Planigale/Sminthopsis)

Dasyurinae

Phascogalini (Phascogale, Antechinus)

Dasyurini (Dasyurus)

Order Peramelemorphia

Peramelidae

Peramelinae (Perameles, Isoodon)

Echymiperinae (Echymipera)

Order Notoryctemorphia

Notoryctidae (Notoryctes)

Order Diprotodontia

Vombatiformes (Vombatus, Phascolarctos)

Macropodiformes

Potoroidae (Aepyprymnus)

Macropodidae (Macropus/Dendrolagus)

Phalangeriformes

Phalangeroidea

Phalangeridae (Phalanger)

Burramyidae (Cercartetus)

Petauroidea

Petauridae (Petaurus)

Pseudocheiridae (Pseudochirops/Pseudocheirus)

Placentalia (Infraclass Eutheria)

Order Primates (Homo)

Order Dermoptera (Cynocephalus)

Order Chiroptera (Pteropus/Tadarida)

Order Artiodactyla (Lama)

Order Eulipotyphla (Talpa/Uropsilus)

Order Perissodactyla (Equus)

Order Proboscidea (Elephas/Loxodonta)

Order Xenarthra (Bradypus)

Commas separate taxa that correspond to distinct terminals in phylogenetic analyses; slashes indicate chimerictaxa that correspond to a single terminal in phylogenetic analyses.

J Mammal Evol

are given in Supplementary Information. These genes were chosen because they havedemonstrated their phylogenetic utility in resolving marsupial interrelationships (e.g., Amrine-Madsen et al. 2003).

PCR products were cleaned using QIAGEN QIAquick PCR purification kits and were thensequenced in both directions at the University of California Riverside’s Core Genetics Institute,which uses an automated DNA sequencer (ABI 3730xl). When necessary, internal sequencingprimers were designed. Accession numbers for previously published sequences and the 19sequences that are new to this study are given in the Supplementary Information.

DNA alignments, data compatibility

New sequences were manually aligned to the Amrine-Madsen et al. (2003) data set after takinginto account amino-acid residues using the program SE-AL (Rambaut 1996). A 63 bp region ofBRCA1 was excluded from phylogenetic analyses following Amrine-Madsen et al. (2003). Apartition homogeneity test (Farris et al. 1994; Swofford 2002) with 1,000 replications and 100taxon input orders per replicate indicated that it was appropriate to combine the five genesegments into one multigene data set (p=0.114: ambiguous regions removed; p=0.113:ambiguous and parsimony uninformative characters removed). In addition, the bootstrapcompatibility method (De Queiroz 1993; Teeling et al. 2002) found no conflicting nodes at90% bootstrap support for the individual genes. The length of the concatenated alignment thatincluded all five gene segments was 6,303 bp (Supplementary Information).

Phylogenetic analyses

Maximum-likelihood (ML) and maximum parsimony (MP) analyses were performed on theconcatenated alignment set using PhyML (Guindon and Gascuel 2003) and PAUP 4.0b10(Swofford 2002), respectively. The best fit model of molecular evolution and associated modelparameters were chosen under the Akaike Information Criterion (AIC) using Modeltest 3.06 forthe ML analyses (Posada and Crandall 1998). Models chosen were TrN+I+Γ (ApoB); GTR+I+Γ (BRCA1, IRBP, Rag1, and concatenation); and TVM+I+Γ (vWF). Heuristic searches using1,000 randomized addition orders with tree-bisection and reconnection (TBR) branch-swappingwere used for MP analyses. The ML analyses were started from a neighbor-joining tree. In allanalyses gaps were treated as missing data. Bootstrap analyses employed the aforementionedoptions with 500 replicates (ML) or 1,000 replicates and 1,000 random input orders (MP).

MrBayes v3.1.1 (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003), whichcarries out Metropolis-coupled Markov chain Monte Carlo sampling, was used to calculateBayesian posterior probabilities. Two Bayesian analyses were performed. In the first Bayesiananalysis, each gene segment in the concatenation was allowed to have its own model ofsequence evolution (models as above). In cases where models were not available underMrBayes (e.g., five substitution types), we selected a more general model (e.g., GTR). In thesecond Bayesian analysis, we used a single model (GTR+Γ+I) for the concatenated data set.We used default settings for priors, random starting trees, and four Markov chains (three hotand one cold). Chains were sampled every 1,000 generations. Analyses were run until theaverage standard deviation of split frequencies for the simultaneous analyses fell below 0.01.

Statistical tests

Alternative phylogenetic hypotheses of phylogenetic relationships were evaluated usingKishino-Hasegawa (KH), Shimodaira-Hasegawa (SH), and approximately unbiased (AU)

J Mammal Evol

statistical tests (Kishino and Hasegawa 1989; Shimodaira and Hasegawa 1999; Shimodaira2002). Each of these tests has disadvantages and can bias tree selection. KH tests can placeoverconfidence in the wrong tree; SH tests can be over conservative; and the AU testscompensate for these tree selection biases although it is only approximately unbiased(Shimodaira 2002). CONSEL was used to perform all three tests (Shimodaira 2002). Weevaluated a priori hypotheses for (1) the root of Marsupialia; (2) the monophyly or paraphylyof Australasian taxa; (3) the placement of Notoryctes; (4) the basal split within Dasyuridae; (5)the monophyly or paraphyly of Phascogalini; (6) the sister group of Echymipera; (7) the sistergroup of Cercartetus; (8) the monophyly or polyphyly of Petauroidea (represented byPetauridae and Pseudocheiridae); and (9) the placement of Macropodiformes.

Molecular dating analyses

We tested the molecular clock hypothesis using the likelihood ratio statistic and it was stronglyrejected (p<0.001) for each of the five genes and for the concatenation. As a result, divergencetimes were estimated using two Bayesian methods that employ a relaxed molecular clock andpermit the incorporation of multiple constraints from the fossil record. The use of multiplefossil constraints provides anchor points throughout the tree, which in turn helps to determinepatterns and degrees of rate variation. Multidivtime (version 9-25-03) (Thorne et al. 1998;Kishino et al. 2001; Thorne and Kishino 2002) assumes autocorrelation of molecular ratesamong lineages, requires a rooted tree topology, and allows for fixed (i.e., hard) minimum andmaximum constraints on selected divergence times. BEAST v.1.4 (Drummond and Rambaut2003; Drummond et al. 2006) simultaneously co-estimates both the phylogeny and divergencetimes, does not assume that lineage rates are autocorrelated, and allows node constraints to betreated as hard or soft (sensu Hedges and Kumar 2004; Yang and Rannala 2006). FollowingYang and Rannala (2006), we plotted the widths of the 95% confidence intervals against themean estimates of divergence times to determine whether the addition of more sequence lengthwill improve the precision of our molecular divergence estimates. A strong linear relationshipsuggests that the addition of more data will have little effect on the precision of the estimates.

Multidivtime

We used the Bayesian phylogeny shown in Fig. 1 for the five-gene concatenation. Branchlengths were estimated using the program estbranches (Thorne et al. 1998; Kishino et al. 2001;Thorne and Kishino 2002); Multidivtime (Thorne et al. 1998; Kishino et al. 2001; Thorne andKishino 2002) was used to estimate divergence times. Two different data sets were createdfrom the five-gene concatenation. In the first data set all of the genes were assumed to changerate by a common factor on each branch, i.e., the concatenation was treated as a single gene. Inthe second data set each gene was allowed gene-specific rate trajectories over time (Thorne andKishino 2002). In both Multidivtime analyses we used the F84 (Swofford et al. 1996) model ofsequence evolution with an allowance for a Γ distribution of rates with four discrete categories.The F84+Γ model was chosen because this is the most complex model implemented inMultidivtime. The transition/transversion parameter and estimates of the rate categories of the Γdistribution were calculated with PAUP 4.0b10 (Swofford 2002) based on the tree shown inFig. 1. We used an age of 75 Ma for the mean of the prior distribution for the root ofMarsupialia, which is 6–10 Ma older than the oldest crown-group metatherian fossils (Case etal. 2005). The mean of the prior distribution for the rate of molecular evolution at the ingrouproot node was set equal to the median amount of evolution from the ingroup root to the ingrouptips divided by the mean of the prior distribution for the root of Marsupialia. Markov Chain

J Mammal Evol

Monte Carlo analyses were run for one million generations after a burnin of 100,000generations to allow Markov chains to approach stationarity before chains were sampled;chains were sampled every 100 generations.

We employed 26 hard constraints based on both the fossil record and previous phylogeneticanalyses for taxa that were included in our analysis. These constraints were as follows (nodenumbers refer to Fig. 2):

(a) Node 7. The 54.6 Ma old Murgon deposit in southeastern Queensland (Godthelp et al.1999) is the oldest terrestrial vertebrate bearing deposit in Australia to produce marsupialtaxa. As of yet no crown group diprotodontians have been recovered. However, since thisis the only Eocene terrestrial mammal bearing deposit in Australia, the absence of fossiltaxa cannot be considered as “hard” evidence that they are not present. The late Oligocenecontains several families of diprotodontians that are morphologically derived in beingsimilar to the living genera. This is in contrast to both the Peramelidae and Dasyuridae.Therefore, we performed two different analyses with different maximums for the base ofDiprotodontia. In the first analysis we used a maximum of 65 Ma, which allows for aslightly earlier origin than suggested by Murgon and in the second analysis we used54.6 Ma. We used 25.5 Ma as the minimum. This is based on the oldest described fossildiprotodontians from Zone A of the Etadunna Formation (see nodes 3 and 4).

Petaurus breviceps

Dromiciops gliroides

Talpidae

Didelphinae

Phascogale tapoatafa

Phalanger orientalis

Caluromys philander

Perameles

Echymipera kalubu

Aepyprymnus rufescens

Pteropus

Vombatus ursinus

Elephantinae

Caenolestes fuliginosus

Phascolarctos cinereus

Monodelphis domestica

Isoodon macrourus

Bradypus

Cercartetus nanus

Pseudocheiridae

Equus

Homo

Macropodidae

Sminthopsinae

Notoryctes typhlops

Antechinus

Dasyurus albopunctatus

Cynocephalus

Lama

Tadarida

Rhyncholestes raphanurus

100100

100100

100100

100100

82.5 51

100100

6946

100100100

10010010065.5

45

100100

100100

9765

100100

82.5 64

100100

100100

100100 100

100

62.5 33

100100

100100

8054 94.5

65 100100

100100

100100

Fig. 1 Bayesian tree with each of the five genes modeled separately for the 6.4kb concatenation. Values aboveand below branches correspond to the mean percentage Bayesian posterior probabilities based on the twosimultaneous runs and the ML bootstrap support percentages, respectively.

J Mammal Evol

(b) Node 1. The Oligo-Miocene deposits of both central Australia and Riversleigh haveproduced several genera of kangaroos. However, there is disagreement over thetaxonomic assignment of these taxa. Case (1984) and Woodburne et al. (1993) treatPurtia mosaicus as a potoroid, Nambaroo species A and B and macropodine genus P sp.A as macropodids. These specimens come from Zone C (25.0–25.5 Ma; Woodburne et al.1993) of the Etadunna Formation. By contrast, other authors treat Nambaroo as abalbarine kangaroo (Long et al. 2002; Cooke 2006) and Purtia as a macropodoid incertaesedis (Long et al. 2002) or even a bulungamayine (Prideaux 2004). These same workersconsider Balbarinae (Balbaridae) the sister group to all other macropodiforms (i.e., stemmacropodiforms) and treat the Bulungamayinae as ancestral to the lophodont kangaroos(e.g., macropodines and sthenurines) even though this group might be paraphyletic. Bycontrast, Flannery (1989) considered the balbarines as ancestral to the lophodontkangaroos. Given the uncertainty in both the taxonomic assignment and placement ofthe Oligo-Miocene fossil taxa we used a conservative minimum of 12 Ma for themacropodid-potoroid split. This minimum is based on the presence of sthenurine-like,postcranial material (Prideaux 2004) and a small, high crowned molar tooth (Murray andMegirian 1992) from the middle Miocene Bullock Creek Local Fauna from the Camfield

Caenolestes fuliginosus

Cercartetus nanus

Phascolarctos cinereus

Isoodon macrourus

Echymipera kalubu

Rhyncholestes raphanurus

Notoryctes typhlops

Vombatus ursinus

Dasyurus albopunctatus

Monodelphis

Dromiciops gliroides

Petaurus breviceps

Didelphinae

Phalanger orientalis

Perameles

Macropodidae

Caluromys philander

Antechinus

Phascogale tapoatafa

Sminthopsinae

Aepyprymnus rufescens

Pseudocheiridae

10.0

L25

L26

L27

L28

L34

L35

0100 80 60 40 20

K-T BoundaryMESOZOIC CENOZOIC

L23

L22

L29

L36

L37

L41

L42

L31

L30

L33

L3

L6

L10

L14

L15

L11

L16

L17

L19

L21

L18

L1

L2

L4

L5

L8

L9

L12

L13

19

20

21

8

14

13

L32

L38

L39

L40

12

AUSTRALIDELPHIA

AMERIDELPHIA

L24

16

14

1111

99

1010

2121

1818

1414

81

1313

1717

1919

2020

111

2

L7

5

37

4

8

16

L2015

666

Fig. 2 Timeline in millions of years before present for marsupial evolution based on the Multidivtime partitionedanalysis (diprotodontian maximum=65Ma). Node numbers refer to those given in Table 5; grey bars indicateconfidence intervals; filled rectangles indicate the oldest known fossil on a given lineage; and fossil constrainednodes are indicated with open circles. L = lineage; K–T = Cretaceous–Tertiary Boundary.

J Mammal Evol

Beds in the Northern Territory (∼12 Ma; Woodburne et al. 1985; Murray and Megirian1992; Long et al. 2002).

Bulungamayines and balbarines are known from Riversleigh’s System A and B deposits(Archer et al. 1999; Cooke 2006), the former of which have been correlated with the NgamaLocal Fauna (Zone D; 24.7–25.0 Ma; Woodburne et al. 1993) of the Etadunna Formationby Myers and Archer (1997). Zone A of the Etadunna Formation has produced a highlyplesiomorphic kangaroo, “Kyeema” (Woodburne et al. 1993), that is not easily assigned toany kangaroo (sub)family. Given that late Oligocene/early Miocene macropodiforms showlittle morphological divergence from each other, and that the oldest kangaroo taxon fromZone A of the Etadunna is highly plesiomorphic and is not clearly associated with either themacropodid or potoroid lineage, we used the base of the Oligocene (33.9 Ma; Gradsteinet al. 2004) as a maximum for the macropodid–potoroid split.

(c) Node 2. Burramyids are known from the late Oligocene to early Miocene deposits of bothcentral Australia and Riversleigh. Burramys wakefieldi is from Zone D of the EtadunnaFormation; Woodburne et al. (1993) proposed a date of approximately 24.7–25.0 Ma forthis zone based on magnetostratigraphy. The burramyid genus Cercartetus is known fromthe late Oligocene Geilston Bay deposits of Tasmania (Tedford and Kemp 1998), which isat least 23 Ma old (Tedford and Kemp 1998). The Geilston Bay deposits have alsoyielded a phalangerid fossil that Tedford and Kemp (1998) classify as Phalangeridaeincertae sedis. Other phalangerids have been described from the middle Miocene ofRiversleigh but their affinities are debated (e.g., Crosby et al. 2004). We used 24.7 Ma asthe minimum and either 54.6 or 65 Ma as the maximum for node 2.

(d) Node 3. The oldest pseudocheirid fossils belong to Paljara sp. A from Zone A(∼25.5 Ma; Woodburne et al. 1993) of the Etadunna Formation. The oldest putativepetaurid fossil is Djaludjangi yadjana from the middle Miocene of Riversleigh(Brammall 1998). However, it has also been suggested that this taxon is Petauroideaincertae sedis (Brammall 1998; Crosby et al. 2004). We used 25.5 Ma as a minimum andeither 54.6 or 65 Ma as the maximum for node 3.

(e) Node 4. The oldest phascolarctid fossils are specimens of Perikoala robusta from Zone Aof the Etadunna Formation, which is approximately 25.5 Ma old following Woodburne etal. (1993). Vombatid fossils belonging to Rhizophascolonus crowcrofti are known from theWipajiri Formation, which unconformably overlies the Etadunna Formation (Woodburneet al. 1993). Undescribed wombats are also known from the late Oligocene of Riversleigh(Archer and Hand 2006). We used 25.5 Ma as a minimum and either 54.6 or 65 Ma as themaximum for node 4.

(f) Node 10. The middle Miocene Barinya wangala of Riversleigh is the oldest fossil that canconfidently be placed into the Dasyuridae (∼11.6–23.0 Ma; Gradstein et al. 2004) (Wroe1998, 2003). These specimens have been referred to their own subfamily (Barinyainae)and are thought to be the sister group to the living subfamilies (Wroe 1998, 2003; Archerand Hand 2006). The oldest dasyurids directly referable to the living subfamilies are fromthe Hamilton Local Fauna, which has been dated at 4.46 Ma (Turnbull et al. 2003). As aresult, we used 4.46 Ma as a minimum for node 10. Dasyurids have been described fromthe Oligo-Miocene Etadunna Formation and several Oligo-Miocene Riversleigh sites butare quite different from living species (Godthelp et al. 1999; Wroe 2003). As a result these“dasyurid” fossils were reassigned to Dasyuromorphia incertae sedis (e.g., Mayigriphus;Godthelp et al. 1999; Wroe 2003) or even Marsupialia incertae sedis (e.g., Wakamatha,Keeuna, Ankotarinja; Godthelp et al. 1999; Wroe 2003). Unfortunately, the next fossilhorizon to produce terrestrial mammals is Murgon (54.6 Ma). Therefore, there is a gap ofalmost 30 Ma in the fossil record, so putting a precise maximum is difficult. Known late

J Mammal Evol

Oligocene to early Miocene dasyurids are archaic. Given this evidence, we use the base ofthe Oligocene (33.9 Ma; Gradstein et al. 2004) as a maximum for the Dasyuridae. Thisallows for the possibility that the late Oligocene to early Miocene dasyurids are directlyrelated to the living dasyurids given that Wroe (1998) diagnoses the Dasyuridae usingfour cranial features not preserved on all of the Oligo-Miocene “dasyurids.”

(g) Node 8. Antechinus sp. from the Hamilton Local Fauna (4.46 Ma; Archer 1982; Turnbullet al. 2003) is the oldest described member of the Phascogalini. We used 4.46 Ma as theminimum and base of the Miocene (23.03 Ma; Gradstein et al. 2004) as the maximum forthe split between Antechinus and Phascogale.

(h) Node 9. Antechinus sp. from the Hamilton Local Fauna (4.46 Ma; Archer 1982; Turnbullet al. 2003) is the oldest described member of either the Dasyurini or Phascogalini. Weused 4.46 Ma as the minimum and base of the Miocene (23.03 Ma; Gradstein et al. 2004)as the maximum for the split between the Dasyurini and Phascogalini.

(i) Node 13. The oldest fossil species referable to crown-group Peramelemorphia is cf.Peroryctes tedfordi from the early Pliocene Hamilton Local Fauna (4.46 Ma; Turnbull etal. 2003). Perameles allinghamensis is known from the Bluff Downs Local Fauna (Archerand Wade 1976). This fauna is derived from the Allingham Formation, which is overlainby basalt dated at 3.62 Ma (Mackness et al. 2000). We used 4.46 Ma as a minimum fornode 13. The oldest described peramelemorphian is Yarala kida from the late OligoceneKangaroo Well local fauna (Schwartz 2006), although a putative perameloid tooth hasbeen reported from Murgon (Archer and Kirsch 2006; Archer and Hand 2006). Yarala isplaced in its own monotypic superfamily and is currently regarded as a stemperamelemorphian (Archer et al. 1999; Archer and Hand 2006). Other fossil bandicootsare known from the late Oligocene to early Pliocene of Riversleigh, the Oligo-MioceneEtadunna Formation, and the Miocene Wipajiri Formation (Woodburne et al. 1993; Archeret al. 1999; Case 2001; Archer and Hand 2006). However, these fossils have not beenformally described. Zone A and B Etadunna Formation bandicoots (25.0–25.7 Ma) arevery different from living bandicoots in having tribosphenic molars (Case 2001). Case(2001) suggests that species from zone D of the Etadunna Formation (24.7–25.0 Ma;Woodburne et al. 1993) and the Wipajiri Formation (10.5–11.5 Ma; Langford et al. 1995)may be ancestral to living forms. The zone D species shows an enlarged metaconule, butretains many of the primitive stem peramelemorphian characters. Two of the three Wipijirispecies are much more derived and appear quite modern but the other species is moresimilar to the zone D species. In view of this evidence, we use the base of the Miocene(23.03 Ma; Gradstein et al. 2004) as the maximum for the Peramelinae–Echymiperinaesplit.

(j) Node 12. The oldest described members of the Peramelinae are Perameles allinghamensis(Bluff Downs Local Fauna; Allingham Formation; Archer and Wade 1976) and Peramelesbowensis (Bow Local Fauna; 3.62 Ma; Muirhead et al. 1997; Mackness et al. 2000). Forthe minimum we used 3.62 Ma and the base of the Miocene (23.03 Ma; Gradstein et al.2004) as the maximum for split between Perameles and Isoodon.

(k) Node 15. The oldest definitive australidelphians come from the Murgon deposits insoutheastern Queensland that have been dated at 54.6 Ma. Murgon fossils include apossible perameloid tooth (Archer and Hand 2006) as well as Marsupialia incertae sedis(Djarthia; Godthelp et al. 1999; Wroe 2003) specimens. We, therefore, used 54.6 as theminimum for the base of Australidelphia. The maximum age for Australidelphia is moredifficult to constrain. The microbiothere Khasia is known from 60.4–59.2 Ma oldBolivian deposits, but there is ongoing debate about the putative microbiothere affinitiesof this genus (see Wroe et al. 2000). Whereas Khasia suggests the possibility of crown-

J Mammal Evol

group australidelphians in the Paleocene, there are no marsupial fossils referable toAustralidelphia from the latest Cretaceous (i.e., Maastrichtian). We therefore used thebase of the Maastrichtian (70.6 Ma; Gradstein et al. 2004) as a cautious maximum for thebase of Australidelphia.

(l) Node 20. Minimum of 12.2 Ma for the split between Didelphidae and Caluromyidaebased on Laventan didelphid fossils described by Marshall (1976). These fossils are fromthe “Monkey Unit” of the Honda Group in Columbia (Marshall 1976) and are at least12.2 Ma old (Czaplewski et al. 2003). Reig et al. (1987, Fig. 69) suggested a late Eocenedate for the basal split between the Caluromyidae and other didelphimorphs. Recentcladistic studies suggest that Paleocene Tiupampan marsupials lie outside of crown groupMetatheria (e.g., Muizon and Cifelli 2001; Sánchez-Villagra and Wible 2002; Horovitzand Sánchez-Villagra 2003; Luo et al. 2003). In light of this evidence we allow that thecaluromyid–didelphid split may have occurred as early as the Paleocene–Eoceneboundary (55.8 Ma; Gradstein et al. 2004) and used this date as a maximum for thebase of Didelphimorphia.

(m) Node 19. Minimum of 6.8 Ma for the Didelphinae–Marmosinae (sensu Kirsch and Palma1995) split based on the occurrence of late Miocene (Huayquerian 9–6.8 Ma) fossilsreferable to both Didelphinae and Marmosinae. Fossils from the medial Miocene(Laventan) have also been referred to Didelphidae (Marshall 1976), possibly to Marmosa(Marshall 1976), Marmosops (Reig et al. 1987), or Micoureus (Goin 1997), but whetherthese forms are stem or crown didelphids remains unclear. More recently, Cozzuol et al.(2006) have described the oldest species of Didelphis from the late Miocene SolimõesFormation in Brazil, which has been assigned to the Huayquerian based onbiochronology but the fossil locality has not been directly dated. Therefore, we used acautious minimum of 6.8 and 55.8 Ma as the maximum for the Didelphinae–Marmosinaesplit.

BEAST

BEAST v1.4 allows for an uncorrelated lognormal distribution (UCLN) model, whichindependently draws the rate of each branch from a lognormal distribution. We performed twodifferent BEAST v1.4 analyses with each of the diprotodontian maxima. In one analysis thegenes were partitioned and in the second analysis the genes were not partitioned.

We used the mean prior distribution for the rate of molecular evolution at the root node ascalculated in the Multidivtime non-partitioned analyses to estimate the mean substitution rateper year in the BEAST non-partitioned analyses. In the partitioned analyses we implementedthe same technique to estimate the mean prior distribution for the rate of molecular evolution atthe root node for each of the individual genes (see “Multidivtime” section). The AkaikeInformation Criterion as implemented in Modeltest 3.06 (Posada and Crandall 1998) was usedto determine the appropriate model of molecular evolution for BEAST analyses. If thesuggested model was nested within the GTR model, the GTR model was used. Branching rateswere drawn from a Yule prior distribution and the starting tree was that shown in Fig. 1.Multiple Markov chain Monte Carlo analyses were run with a burnin equivalent to the first10%. These independent runs were then combined (30–40 million generations in total) until theestimated sample size was at least 100. The chains were sampled every 1,000 generations.

Tracer 1.2 (Rambaut and Drummond 2003) was used to visually check for mixing/stationarity. Bounded ranges were used to calibrate the molecular divergence estimates. Weemployed hard-bounded constraints that were compatible with those used in the Multidivtime

J Mammal Evol

analyses to allow for direct comparisons. We also explored the affect of using soft bounds. Forthese analyses, the prior node constraint followed a standard normal distribution with 95% ofthe distribution between the upper and lower bounds and 2.5% of the distribution in each tail.Minimum and maximum constraints on divergence times for BEAST analyses were the sameas those that were used in Multidivtime analyses (above).

Unrepresented basal branch lengths

We estimated the fraction of the marsupial fossil record that is missing using methodsdescribed in Teeling et al. (2005). We compiled a list consisting of every definitive fossil forevery marsupial lineage present on our phylogeny (Fig. 1). We attempted to use unequivocalfossils known from more than one element, but sometimes this was not possible (Table 2). Alineage is defined to include any taxon on that branch as well as any “off-shoots” from thatbranch (22 terminal taxa and 43 internal branches). The oldest fossil representative for eachsister-lineage pair was then compared to the corresponding Bayesian molecular date estimate todetermine the unrepresented basal branch length (UBBL) for each branch (Teeling et al. 2005).No missing data (UBBL=0) were recorded for a given lineage when the molecular and fossilage estimates were the same or if the molecular age estimate was younger than the fossilestimate. If the fossil age estimate was younger than the molecular age estimate, UBBL wasdetermined by subtracting the age estimate of the fossil from the molecular age estimate.

Results

Phylogenetic analyses

Figure 1 shows the average posterior probabilities for the two Bayesian analyses that allowedeach gene segment to have its own model of sequence evolution and the ML bootstrap supportpercentages. Table 3 summarizes posterior probabilities for Bayesian analyses and bootstrapsupport percentages for ML and MP analyses.

The root of Marsupialia was not well resolved, although a basal split betweenDidelphimorphia versus Australidelphia + Paucituberculata was favored in Bayesian and MLanalyses. Bayesian posterior probabilities for Australidelphia + Paucituberculata ranged from0.82 to 0.89; the ML bootstrap support for this clade was 64%. The monophyly ofAmeridelphia was not well supported (posterior probabilities=0.06; ML bootstrap support=13.6; MP bootstrap support=0), although the American orders Didelphimorphia andPaucituberculata were each monophyletic in all analyses. Within Didelphimorphia, didelphidsgrouped to the exclusion of Caluromys (posterior probabilities=1.00; ML and MP bootstrapsupport=100).

Australidelphia was strongly supported in Bayesian (posterior probabilities=1.00), ML(bootstrap support=100), and MP (bootstrap support=100) analyses. Within Australidelphia,Bayesian and ML analyses also support the monophyly of Australasian taxa (Diprotodontia,Peramelemorphia, Dasyuromorphia, Notoryctemorphia) to the exclusion of the SouthAmerican order Microbiotheria (posterior probabilities=0.69–0.87; ML bootstrap=46; MP=43). Within the Australasian moiety, Peramelidae + Notoryctes + Dasyuridae formed a clade tothe exclusion of Diprotodontia in Bayesian and ML analyses (posterior probabilities=0.97–0.99; ML bootstrap support=65; MP=51.0). All of the non-monotypic Australasian orderswere recovered as monophyletic with strong support (posterior probabilities and bootstrapsupport of 1.00 and 100, respectively). Within Peramelidae, Isoodon and Perameles grouped to

J Mammal Evol

Tab

le2

Estim

ationof

themissing

fossilrecord

perlin

eage

BranchNum

ber

Taxon

OldestFossila

Molecular

Age

Duration

ofLineage

TotalMissing

perLineage

Percent

Missing

perLineage

Reference

AUSTRALID

ELPHIA

Diprotodontia

Macropodiform

es

L1

Bettongia

moyesib

5.33

13.8/16.1

13.8/16.1

8.5/10.77

61.5/66.9

Woodburne

etal.1993

L2

Macropodine

gen.

Psp.A

b12.0

13.8/16.1

13.8/16.1

1.8/4.1

13.2/25.5

Woodburne

etal.1993

L3

Riversleigh

Balbaridaec

25.25

44.9/45.6

31.1/29.5

5.8/4.3

18.8/14.4

Cooke

andKear1999

Phalangeroidea

L4

Burramys

wakefieldi

24.85

41.9/45.5

41.9/45.5

17.0/20.6

40.1/45.3

Woodburne

etal.1993

L5

GeilstonBay

phalangerid

23.00

41.9/45.5

41.9/45.5

18.9/22.4

45.1/49.3

TedfordandKem

p1998

L6

Unknown

44.9/45.6

3.0/0.2

3.0/0.2

100/100

L7

Unknown

47.6/45.6

2.7/0.0

2.7/0.0

100/100

Petauroidea

L8

Riversleigh

petaurid

24.85

33.1/28.4

33.1/28.4

8.2/3.6

24.9/12.5

BrammallandArcher1999

L9

Paljara

sp.A

25.60

33.1/28.4

33.1/28.4

7.5/2.8

22.6/9.9

Woodburne

etal.1993

L10

dPilkipild

rataylori

25.60

47.6/45.6

14.5/17.2

14.5/17.2

100/100

Woodburne

etal.1993

L11

Unknown

54.1/53.8

6.4/8.2

6.4/8.2

100/100

Vom

batiformes

L12

Perikoala

robusta

25.60

37.3/31.7

37.3/31.7

11.7/6.1

31.4/19.2

Woodburne

etal.1993

L13

EtadunnaV

ombatoids

25.60

37.3/31.7

37.3/31.7

11.7/6.1

31.4/19.2

Woodburne

etal.1993

L14

Unknown

54.1/53.8

16.8/22.1

16.8/22.1

100/100

L15

Unknown

62.2/62.8

8.1/9.0

8.1/9.0

100/100

Dasyuromorphia

Dasyuridae

Phascogalini

L16

eAntechinussp.

4.46

5.7/7.7

5.7/7.7

1.3/3.2

22.2/42.1

Archer1982;Turnbull

etal.2003

J Mammal Evol

Tab

le2

(contin

ued)

BranchNum

ber

Taxon

OldestFossila

Molecular

Age

Duration

ofLineage

TotalMissing

perLineage

Percent

Missing

perLineage

Reference

L17

Phascogalesp.

0.036

5.7/7.7

5.7/7.7

5.7/7.7

99.4/99.5

Pledge1990

L18

Unknown

10.2/9.9

4.4/2.2

4.4/4.4

100/100

Dasyurini

L19

fGanbulanyidjadjin

guli

13.79g

10.2/9.9

10.2/9.9

0.0/0.0

0.0/0.0

Wroe1998

L20

Unknown

15.0/22.9

4.9/13.0

4.9/13.0

100/100

Sminthopsini

L21

hPlanigale

sp.

4.4

15.0/22.9

15.0/22.9

10.6/18.5

70.8/80.8

Archer1982

L22

iAnkotarinja

tirarensis,

Keeunawoodburnei

25.25

59.1/59.9

44.0/37.0

18.8/11.8

42.7/31.8

Woodburne

etal.1993

Notoryctemorphia

L23

jRiversleigh

notoryctid

19.50

59.1/59.9

59.1/59.9

39.6/40.4

67.0/67.5

Archeret

al.1999

;Warburton

2003

L24

Unknown

60.9/60.1

1.8/0.2

1.8/0.2

100/100

Peram

elem

orphia

Peram

elidae

L25

kPeroryctestedfordi

4.46

9.1/13.3

9.1/13.3

4.6/8.8

51.0/66.5

Turnbullet

al.2003

L26

Isoodonobesulus

0.036

4.5/6.7

4.5/6.7

4.4/6.7

99.2/99.5

Price

2004

L27

Peram

eles

allin

gham

ensis

4.4

4.5/6.7

4.5/6.7

0.1/2.3

1.4/34.3

ArcherandWade1976

L28

Unknown

9.1/13.3

4.6/6.6

4.6/6.6

100/100

L29

lYarala

kida

23.03

60.9/60.1

51.8/46.8

28.8/23.8

55.5/50.8

Schwartz

2006

L30

Unknown

62.2/62.8

1.3/2.7

1.3/2.7

100/100

L31

Unknown

62.9/62.8

0.7/0.0

0.7/0.0

100/100

Microbiotheria

L32

Khasiacordillerensis

59.80

62.9/62.8

62.9/62.8

3.1/3.0

4.9/4.8

MarshallandMuizon1988

;Marshallet

al.1997

L33

mChulpasia

mattaueri

56.60

74.8/85.6

12.0/22.8

12.0/22.8

100/100

Goinet

al.2006

;Sigé

etal.2004

AMERID

ELPHIA

Paucituberculata

L34

Unknown

9.7/14.1

9.7/14.1

9.7/14.1

100/100

L35

Unknown

9.7/14.1

9.7/14.1

9.7/14.1

100/100

J Mammal Evol

Tab

le2

(contin

ued)

BranchNum

ber

Taxon

OldestFossila

Molecular

Age

Duration

ofLineage

TotalMissing

perLineage

Percent

Missing

perLineage

Reference

L36

nMayulestesferox

59.80

74.8/85.6

65.2/71.5

5.4/11.7

8.3/16.4

Goinet

al.2006

L37

Unknown

78.5/89.3

3.6/3.7

3.6/3.7

100/100

Didelphim

orphia

L38

Calurom

ysderbianus

0.01

37.5/35.4

37.5/35.4

37.5/35.4

100/100

Reiget

al.1987

L39

oHyperdidelphispattersoni

7.90

27.9/25.0

27.9/25.0

20.0/17.1

71.7/68.4

Reig1952

,1958;Reiget

al.

1987

L40

Micoureus

laventicus

12.85

27.9/25.0

27.9/25.0

15.0/12.2

53.9/48.6

Goin1997

L41

Unknown

37.5/35.4

9.7/10.4

9.7/10.4

100/100

L42

pNortedelphysinterm

edius

67.0

78.5/89.3

41.0/59.3

11.5/22.3

28.0/41.2

Goinet

al.2006

Analysesbefore

andafterslashesarebasedon

thepartition

edhard-bou

nded

Multid

ivtim

eandBEASTanalyses

(Diprotodontiamaxim

umsetto65

Ma),respectively.Valuesarein

millions

ofyears.Referencesreferto

column3(age

ofoldestfossil).

aIfpo

ssible

only

taxa

from

dateddepo

sitswereused.Ifarang

ewas

given,

themidpointdate/value

was

used

forcolumn3(age

ofoldestfossil).

bAspreviously

mentio

nedthereis

disagreementin

regardsto

theexactidentificationandaffinitiesof

theOlig

o-Miocene

macropodiform

s.For

thepurposes

oftheUBBL

analyses

wetreatedallof

theOlig

o-Miocene

kangaroo

speciesas

stem

macropodoids.

cMyersandArcher(199

7)have

correlated

oneof

theoldestRiversleigh

sitesto

theNgamaLocalFauna

oftheEtadunnaFormation(24.7–25

.0Ma;Woo

dburne

etal.1

993).W

eused

themiddleof

theestim

ated

ageof

theNgamaLocal

Fauna

forallRiversleigh

System

Adeposits.

dAlth

ough

considered

incertae

sedisby

Archeret

al.(198

7)or

astem

lineage

oftheph

alangeroids(M

arshallet

al.19

90),thepilkipild

rids

aresometim

esconsidered

thesister

grou

pof

thepetauroids.

eWeassigned

aconservativ

edateas

theendof

thePleistocene.Inaperson

alcommunicationto

ArcherandHand(200

6),v

anDyckhassuggestedthataphascogalin

esimilarto

Antechinu

smight

bepresentin

earlyMiocene

ofRiversleigh

.How

ever,ithasyetto

bedescribedandthisassign

mentcanbe

view

edas

high

lyqu

estio

nable.

fThisspeciescomes

from

themiddleto

earlylateMiocene

EncoreSite

inRiversleigh.T

here

isdisputein

regardsto

exactdatesatRiversleigh

(e.g.,Megirianetal.2

004).A

sa

resultweusethemiddleof

themedialMiocene

(Gradstein

etal.20

04)as

reasonable

approxim

ation.

gIn

both

theBEASTandMultid

ivtim

eanalyses,the

ageof

thefossilwas

olderthan

thedivergence

estim

ates.G

anbu

lanyim

ight

also

bemorecloselyrelatedto

Barinya,w

hich

isconsidered

thesister

taxonto

theliv

ingdasyurid

subfam

ilies

(Wroe20

03),which

isconsistent

with

ouryo

ungerageestim

atefortheDasyurini

Phascog

alinisplit.

hThe

minim

umageof

theBluffsDow

nLocalFauna

(Allingham

Formation)

is3.62

Mawith

amaxim

umof

5.2Mabasedon

potassium-argon

basaltdates(M

acknessetal.2

000;

ArcherandWade19

76).

iGod

thelpet

al.(199

9)andWroe(200

3)remov

ethesegenera

from

Dasyuridaeandreferto

them

asMarsupialia

incertae

sedis.Evenifthesegenera

laterturn

outno

tto

bemem

bersof

theDasyuromorphia,thylacinidfossils

areknow

nfrom

thelateOlig

ocene(W

roe20

03)andmolecular

studieshave

demonstratedasister

grouprelatio

nshipwith

the

J Mammal Evol

Dasyuridae(K

rajewskiet

al.19

97).TedfordandKem

p(199

8)describe

theproxim

alendof

afemur

from

thelate

Olig

oceneGeilstonBay

localfauna,

which

they

referto

the

Dasyuridae.

How

ever,sinceon

lyon

eelem

entiskn

ownandbecauseof

prob

lemswith

morph

ological

conv

ergence,

thisassign

mentcanon

lybe

regarded

astentative.

jNotoryctid

specim

ensarekn

ownfrom

severalRiversleigh

System

Blocalitiesthatarethou

ghtto

beearlyto

middleMiocene

inage,bu

tthesehave

yetto

bewellcorrelated

toanydatedfossildeposit.Asaresultweused

themiddleMiocene

datesof

Gradstein

etal.(200

4).

kAssignm

entto

Perorycteswas

considered

prov

isionalby

Turnb

ullet

al.(200

3).

lThese

specim

ensarefrom

theKangarooWellLocal

Fauna

which

hasnotbeen

directly

dated,

butMegirianet

al.(200

4)suggestalate

Olig

oceneagebasedon

biocorrelatio

n.Theysuggestthatthefaun

ais“slig

htly”olderthan

theWipajirifauna(11.5–10

.5Ma;Langfordetal.1

995)

but“slig

htly”yo

ungerthan

Zon

eDof

theEtadunn

aFormation(25–

24.7Ma;Woo

dburne

etal.1

993).A

saresultweused

theOlig

o-Miocene

boundary

(23.03

Ma;Gradstein

etal.2

004)

asan

ageestim

ateforthisspecim

en.Itsho

uldbe

notedthat

apu

tativ

e“peram

eloid”

toothisknow

nfrom

Murgon,

buthasyetto

bedescribed(A

rcherandHand20

06),which

would

extend

therecord

back

totheEocene.

mChulpasia

from

theLateCretaceousor

Paleocene

UmayoFormationof

southern

Peruhasbeen

associated

with

Glasbius(Crochet

andSigé19

93),Thylacotin

gafrom

the

Murgo

ndepositsof

Australia(Sigéetal.1

995),and

inthemostrecent

studysister

totheMicrobiotheria(G

oinetal.2

006).A

saresultwetreatitas

anaustralid

elphianin

allof

thefossilcompletenessestim

ates.T

heUmayoFormationisamicroconglomeratethatisnotage

constrained,

butn

ewpaleontological,paleom

agnetic,and

stratig

raphicdatafavor

chron24

(latestPaleocene

orearliestEocene)

over

chrons

26and29

(Sigéetal.2

004).A

saresultweusethebase

ofchron24

(56.6Ma)

asits

age.AcloselyrelatedChulpasia

sp.may

also

bepresentin

Murgo

n(Sigéet

al.20

04).

nKielan-Jaworow

skaet

al.(200

4)treattheLancian

Glasbiid

aeas

afamily

with

inPaucituberculata.

How

ever,themostrecent

cladistic

analyses

dono

tfind

anyassociation

betweenGlasbiusandthepaucitu

berculates

(Goinet

al.20

06).Goinet

al.(200

6)recoveredtheBorhyaenoidea

asthesister

groupto

paucitu

berculates.How

ever,itshould

benotedthatSánchez-Villagra

etal.(20

07)recovertheBoryh

yaenidae

asou

tsideof

theliv

inggenera

andotherearlierstud

ies(e.g.,Rougier

etal.1

998;

Luo

etal.2

003)

sugg

estthe

Boryhyaenidae

arenotcrow

ngroupmarsupialsas

well.

oReigetal.(19

87)foun

dacloseassociationbetweenHyperdidelphisandDidelph

is.O

therwisetheoldestmem

berof

thislin

eage

isLutreolinatracheia

from

thelateMiocene

ofArgentin

a(Sim

pson

1974)or

Didelphissolim

oensisfrom

thelate

Miocene

Solim

öesFormation(Cozzuol

etal.20

06).

pGoinet

al.(200

6)recoveredtherecently

named

Lancian

genusNortedelphys(H

erpetotheriid

ae)as

thesister

taxonto

allotherdidelphimorphians.How

ever,someworkers

suggestthereareno

crow

ngroupmetatherianspresent/k

nownin

theCretaceous(e.g.,Rougier

etal.19

98;Luo

etal.20

03;Sánchez-Villagra

etal.20

07).

J Mammal Evol

Table 3 Summary of bootstrap and posterior probabilities

Hypothesis MP ML Bayesian analyses

Partitioned Non-partitioned

Run 1 Run 2 Run 1 Run 2

Ameridelphia (Didelphimorphia + Paucituberculata) 0 13.6 0.06 0.06 0.02 0.03

Australidelphia 100 100 1.00 1.00 1.00 1.00

All marsupials except Dromiciops 0 0 0.00 0.00 0.00 0.00

All marsupials except Paucituberculata 95 22.0 0.11 0.12 0.09 0.07

All marsupials except Didelphimorphia 4.9 64.4 0.83 0.82 0.89 0.89

Monophyly of Australasian taxa 43.2 46.2 0.69 0.69 0.87 0.87

Eometatheria (All Australasian taxa but Peramelidae) 0 0 0.00 0.00 0.00 0.00

Australasian possum monophyly 26.2 6.2 0.001 0.00 0.00 0.00

Notoryctes + Diprotodontia 0 0 0.00 0.00 0.00 0.00

Notoryctes + Dasyuridae 39.2 44.6 0.66 0.65 0.67 0.68

Notoryctes + Peramelidae 32.8 16 0.08 0.09 0.06 0.06

Peramelidae + Notoryctes + Dasyuridae 51.0 64.6 0.97 0.97 0.99 0.99

Peramelidae + Notoryctes + Dasyuridae + Dromiciops 31.5 9.2 0.04 0.04 0.00 .00

Peramelidae + Dasyuridae 27.4 35.0 0.26 0.26 0.27 0.26

Peramelidae + Notoryctes + Dromiciops 0 1.4 0.00 0.00 0.00 0.00

Peramelidae + Dasyuridae + Diprotodontia 0 0 0.00 0.00 0.00 0.00

Dromiciops + Diprotodontia 0 23.4 0.25 0.26 0.11 0.11

Dromiciops + Dasyuridae 30.9 6.8 0.00 0.00 0.00 0.00

Dromiciops + Peramelidae 12.0 7.8 0.00 0.00 0.00 0.00

Dromiciops + Dasyuridae + Peramelidae 0 0.2 0.00 0.00 0.00 0.00

Dromiciops + Dasyuridae + Notoryctes 0 10.2 0.02 0.01 0.00 0.00

Diprotodontia + Peramelidae 0 0 0.00 0.00 0.00 0.00

Paucituberculata 100 100 1.00 1.00 1.00 1.00

Didelphimorphia 100 100 1.00 1.00 1.00 1.00

Didelphinae + Marmosinae 100 100 1.00 1.00 1.00 1.00

Dasyuridae 100 100 1.00 1.00 1.00 1.00

Phascogalini 100 100 1.00 1.00 1.00 1.00

Dasyurini + Phascogalini 100 100 1.00 1.00 1.00 1.00

Peramelidae 100 100 1.00 1.00 1.00 1.00

Echymiperinae 100 100 1.00 1.00 1.00 1.00

Diprotodontia 100 100 1.00 1.00 1.00 1.00

Vombatiformes 100 100 1.00 1.00 1.00 1.00

Macropodiformes + Phalangeriformes 100 100 1.00 1.00 1.00 1.00

Burramyidae + Macropodiformes 2.4 1.2 0.00 0.00 0.00 0.00

Macropodiformes + Phalangeridae 10.2 4.2 0.00 0.00 0.00 0.00

Macropodiformes 100 100 1.00 1.00 1.00 1.00

Macropodiformes + Phalangeroidea 11.6 50.8 0.82 0.83 0.74 0.74

Petauroideaa 100 100 1.00 1.00 1.00 1.00

Macropodiformes + Petauroidea 48.2 38.2 0.18 0.17 0.26 0.26

Phalangeroidea 72.5 88.8 1.00 1.00 1.00 1.00

Partitioned each gene was partitioned to have its own model of molecular evolution, Non-partitionedconcatenation was treated as a single gene, MP maximum parsimony, ML maximum likelihooda Represented by Petauridae and Pseudocheiridae

J Mammal Evol

the exclusion of Echymipera. Within Dasyuridae, Phascogalini (Antechinus and Phascogale)was monophyletic and Phascogalini and Dasyurini (Dasyurus) grouped to the exclusion ofSminthopsinae. Among diprotodontian marsupials, Vombatiformes and Macropodiformes +Phalangeriformes were reciprocally monophyletic clades. Macropodiformes was monophyleticwhereas there was no support for themonophyly of Phalangeriformes. However, there was supportfor two distinct Australasian possum clades: a petauroid clade represented by Pseudocheiridae +Petauridae, and a phalangeroid clade represented by Phalangeridae and Burramyidae.

Statistical tests

Table 4 reports the results of the KH, SH, and AU tests. Four hypotheses were compared forthe root of Marsupialia. The pairwise comparisons rejected rooting the tree betweenDromiciops gliroides and other marsupials (Hershkovitz 1992) and rooting betweenAmeridelphia and Australidelphia. However, these tests could not discriminate betweenrooting the marsupial tree on Didelphimorphia or on Paucituberculata. No significantdifferences were found between Eometatheria (Dromiciops gliroides + all Australasian ordersexcepting Peramelemorphia; Kirsch et al. 1997); a monophyletic Australasian clade; and aDromiciops + Diprotodontia clade as recovered by Drummond et al. (2006).

Four hypotheses were compared for the placement of Notoryctes. The pairwise comparisonsrejected a sister group relationship between Notoryctes and the Diprotodontia, but failed todiscriminate between a sister-taxon relationship to either the Dasyuridae, Peramelidae, orDasyuridae + Peramelidae. Within Dasyuromorphia, we compared two different treehypotheses for both the base of Dasyuridae and the sister taxon to Antechinus. There wasstatistical support for a basal split between the Sminthopsinae and other dasyurids as found byKrajewski et al. (2000) rather than a basal split between the Dasyurini and other dasyurids(Wroe et al. 2000). Phascogalini monophyly was also statistically supported, contrary to theresults of Wroe et al. (2000) and consistent with Krajewski et al. (2000). Within Peramelidae,there was statistical support for an Echymipera + Peramelinae clade as compared to a sistergroup relationship to either Isoodon or Perameles.

Within Diprotodontia, there was statistical support for the Petauroidea (represented byPetaurus and Pseudocheiridae) as compared to a dichotomous split between either Petaurus(Horovitz and Sánchez-Villagra 2003) or the Pseudocheiridae (Woodburne et al. 1987) and theother diprotodontians.

We compared five possible positions for Cercartetus (Family Burramyidae). These testsrejected a sister-group relationship to all other diprotodontians except Petaurus (Horovitz andSánchez-Villagra 2003) and a sister-group relationship to the Macropodiformes (rejected in alltests but the SH), but were unable to discriminate between a sister-taxon relationship to eitherthe Petauroidea (Springer and Woodburne 1989), the Phalangeridae (Springer and Kirsch1991), or a Phalangeridae + Macropodiformes clade. However, the tests favored monophyly ofthe Phalangeroidea.

Statistical tests failed to discriminate between three alternative hypotheses for the sistergroup of the Macropodiformes, i.e., Phalangeroidea (Springer and Kirsch 1991), Petauroidea(Amrine-Madsen et al. 2003), and Phalangeriformes (possum monophyly; Springer andWoodburne 1989), but favored the Phalangeroidea + Macropodoidea hypothesis.

Molecular dating

Figure 2 shows a timescale for marsupial divergences based on the results of a Multidivtimeanalysis in which each gene was allowed to have its own model of sequence evolution

J Mammal Evol

Table 4 Results of statistical tests

Phylogenetic hypotheses -ln likelihood Δ P

KH SH AU

Base of Marsupialia

(a) Dromiciops and other marsupials 63840.24996 49.64288 0.000* 0.000* 0.000*

(b) Paucituberculata and other marsupials 63793.16680 2.55972 0.259 0.587 0.261

(c) Didelphimorphia and other marsupials (best) 63790.60708 0.741 0.868 0.744

(d) Ameridelphia and Australidelphia 63842.64640 52.03932 0.000* 0.000* 0.000*

Australasian taxa

(a) Monophyletic (best) 63790.60708 0.656 0.797 0.699

(b) Eometatheria 63795.65829 5.05121 0.344 0.593 0.428

(c) Dromiciops + Diprotodontia 63792.32610 1.71902 0.262 0.329 0.273

Position of Notoryctes (sister taxon to)

(a) Dasyuridae (best) 63790.60708 1.000 0.745 0.739

(b) Peramelidae 63792.87153 2.26445 0.505 0.612 0.261

(c) Diprotodontia 64074.21936 283.61228 0.000* 0.000* 0.000*

(d) Dasyuridae + Peramelidae (best) 63790.60708 0.000 1.000 0.745 0.739

Base of Dasyuridae

(a) Sminthopsinae (best) 63790.60708 1.000 1.000 1.000

(b) Dasyurini (Dasyurus) 63863.15502 72.54794 0.000* 0.000* 0.000*

Antechinus (sister taxon to)

(a) Phascogale (Phascogalini monophyly) (best) 63790.60708 1.000 1.000 1.000

(b) Sminthopsinae 63951.07986 160.47278 0.000* 0.000* 0.000*

Position of Echymipera (sister taxon to)

(a) Peramelinae (best) 63790.60708 1.000 1.000 1.000

(b) Isoodon 63854.20520 63.59812 0.000* 0.000* 0.000*

(c) Perameles 63853.91710 63.31002 0.000* 0.000* 0.000*

Position of Cercartetus (sister taxon to)

(a) Phalangeridae (best) 63790.60708 0.936 0.983 0.914

(b) Petauroideaa 63802.47924 11.87216 0.132 0.456 0.143

(c) all other Diprotodontia 64035.67588 245.06880 0.000* 0.000* 0.000*

(d) Macropodiformes 63802.12219 11.51510 0.039* 0.439 0.024*

(e) Macropodiformes + Phalangeridae 63800.86253 10.25544 0.064 0.473 0.089

Petauroideaa (monophyly)

(a) Petaurus + Pseudocheiridae (best) 63790.60708 1.000 1.000 1.000

(b) Petaurus sister to other Diprotodontia 64001.15798 210.55090 0.000* 0.000* 0.000*

(c) Pseudocheiridae sister to other Diprotodontia 63922.33414 131.72706 0.000* 0.000* 0.000*

Position of Macropodiformes (sister group to)

(a) Phalangeroidea (best) 63790.60708 0.596 0.718 0.637

(b) Petauroideaa 63792.06444 1.45736 0.404 0.543 0.462

(c) Phalangeriformes (possum monophyly) 63796.59513 5.98805 0.139 0.221 0.073

KH Kishino-Hasegawa, SH Shimodaira-Hasegawa, AU approximately unbiased

*P=<0.05a Represented by Petauridae and Pseudocheiridae

J Mammal Evol

(partitioned) and the maximum age for the base of Diprotodontia was fixed at 65 Ma. Table 5shows point estimates of divergence times, 95% credibility intervals (Multidivtime), and 95%highest posterior densities (HPDs) for Multidivtime and hard-bounded BEAST analyses for allnodes in Fig. 2. Although the maximum of 33.9 Ma (nodes 1 and 10) can be consideredarbitrary, the 95% HPDs and 95% credibility intervals are not bumping up against themaximum. As a result, the upper bound is not critical in informing these divergence dates.This is in contrast to the diprotodontian maximum, which is critical in informing thedivergence estimates in that the 95% HPDs and 95% credibility intervals are bumping upagainst the maximum. In analyses that employed the 65 Ma maximum for the base ofDiprotodontia, mean divergence dates obtained with Multidivtime were 1.5 Ma younger thanthose obtained with BEAST. In analyses that employed the 54.6 Ma maximum for the base ofDiprotodontia, mean BEAST dates were 2.4 Ma older than Multidivtime dates. Thediprotodontian maximum had less influence on the BEAST dates than on the Multidivtimedates. Mean BEAST dates were 0.87 Ma older with the 65 Ma maximum whereas meanMultidivtime dates were 1.8 Ma older.

Results of Multidivtime and BEAST analyses that employed a single model of sequenceevolution for the entire concatenation of gene sequences (i.e., non-partitioned) are similar toresults obtained with partitioned analyses and are provided in the Supplementary Information.Mean dates in the non-partitioned analyses (54.6 Ma diprotodontian maximum) were slightlyolder (≤1 Ma), but when the diprotodontian maximum was 65 Ma they were slightly younger(≤1 Ma) than the corresponding dates in the partitioned analyses. The results of the soft-bounded BEAST analyses (Supplementary Information) were similar to corresponding resultsthat were obtained with hard bounds. Mean dates were ≤1 Ma younger in all analyses(Supplementary Information).

All point estimates (74–89 Ma), 95% credibility intervals, and 95% HPDs for the base ofMarsupialia were in the Cretaceous (Table 5). Point estimates were also entirely in theCretaceous for the split between Paucituberculata and Australidelphia (71–86 Ma). The mostrecent common ancestor of Australidelphia (node 16) was estimated at 60–66 Ma. Interordinaldivergences within the Australasian clade (nodes 11, 14, 15) ranged from 56 to 64 Ma. Allintraordinal divergence dates were placed in the Cenozoic. The deepest intraordinal splits werewithin Diprotodontia.

Figure 3 shows the widths of the 95% confidence intervals plotted against the meanestimates of divergence times for the hard-bounded Multidivtime analysis (Diprotodontiamaximum set to 65 Ma). The Multidivtime analyses yielded a more linear relationship ascompared to the BEAST analyses (other Multidivtime analyses and BEAST analyses notshown). The nearly linear relationship obtained in the Multidivtime analyses suggests that theaddition of more sequence data will not significantly improve our Multidivtime estimated timesof divergence. Put differently, the regression line for the partitioned Multidivtime analyses (y=0.2725x) suggests that even with an infinite amount of sequence data each 1 Ma of speciesdivergence will only add 0.2725 Ma to the 95% confidence interval. However, this relationshipwill only hold up if the constraints remain unchanged. BEAST plots (not shown) were non-linear, which suggests that more sequence data will improve estimates of divergence timeobtained with BEAST.

After the exclusion of the burnin trees, the non-partitioned BEAST analyses recovered 65–80 different tree topologies with the MAP (maximum a posteriori) tree accounting for 64–65%of the posterior probability. These tree topologies for Marsupialia were identical to that of theMrBayes tree shown in Fig. 1. The covariance statistic indicated that there was virtually noautocorrelation of rates for all of the non-partitioned analyses (Supplementary Information)with a mean rate of evolution equal to 1.3E-3–1.4E-3 substitutions per site per million years

J Mammal Evol

(HPD=1.3E-3–1.5E-3). The partitioned BEAST analyses recovered thousands of unique treetopologies with the MAP tree for each gene accounting for less than 1–14% of the posteriorprobability. The mean of the covariance statistics for BRCA1, IRBP, and vWF in some of theanalyses indicates that there was some autocorrelation of rates for these genes; ApoB and Rag1showed no evidence of autocorrelation (Supplementary Information).

Unrepresented basal branch lengths

Tables 2 and 6 show the percentage of the UBBL for the lineages and clades based on thepartitioned, hard-bounded Multidivtime and BEAST analyses (Diprotodontia maximum=65 Ma). External lineages on the marsupial tree were on average 29% (BEAST) to 32%(Multidivtime) more complete than internal lineages (Table 6). In comparison to the Felidae and

Table 5 Hard bounded partitioned BEAST and Multidivtime divergence estimates

Nodeb Divergence Estimatesa

Multidivtime BEAST

Diprotodontia maximum

54.6Ma 65Ma 54.6Ma 65Ma

Macropodiformes (node 1) 13.4 (12.1–15.8) 13.8 (12.1–17.0) 16.0 (12.0–22.2) 16.1 (12.0–23.3)

Phalangeroidea (node 2) 39.8 (35.5–44.0) 41.9 (36.0–48.2) 43.4 (36.3–50.7) 45.4 (36.4–54.6)

Petauridae + Pseudocheiridae (node 3) 31.4 (27.2–35.7) 33.1 (27.7–38.9) 28.0 (25.5–32.5) 28.4 (25.5–34.0)

Vombatiformes (node 4) 35.4 (31.0–39.7) 37.3 (31.4–43.6) 30.8 (25.5–38.3) 31.7 (25.5–40.7)

Macropodiformes + Phalangeroidea(node 5)

42.7 (38.4–46.7) 44.9 (38.9–51.3) 43.7 (37.0–50.1) 45.6 (36.8–54.5)

Macropodiformes + Petauroideac +Phalangeroidea (node 6)

45.2 (40.9–49.0) 47.6 (41.4–54.1) 43.8 (37.0–50.1) 45.6 (36.8–54.5)

Diprotodontia (node 7) 51.3 (46.8–54.4) 54.1 (47.3–60.9) 50.8 (45.8–54.6) 53.8 (44.6–62.8)

Phascogalini (node 8) 5.5 (4.5–7.0) 5.7 (4.6–7.5) 7.8 (4.5–11.3) 7.7 (4.5–11.4)

Dasyurini + Phascogalini (node 9) 9.7 (7.9–12.0) 10.2 (8.1–12.8) 10.0 (6.4–14.5) 9.9 (5.8–14.4)

Dasyuridae (node 10) 14.4 (12.0–17.2) 15.0 (12.3–18.3) 22.6 (16.6–29.3) 22.9 (16.6–29.7)

Notoryctes + Dasyuridae (node 11) 56.2 (51.2–61.3) 59.1 (51.7–66.4) 59.1 (51.9–67.1) 59.9 (52.6–67.9)

Peramelinae (node 12) 4.3 (3.6–5.5) 4.5 (3.6–5.8) 6.8 (3.6–11.2) 6.7 (3.6–10.8)

Peramelidae (node 13) 8.7 (7.1–10.6) 9.1 (7.3–11.3) 13.7 (7.7–20.1) 13.3 (7.2–20.1)

Peramelidae + Notoryctes +Dasyuridae (node 14)

58.0 (53.1–62.8) 60.9 (53.5–68.1) 59.3 (52.3–66.9) 60.1 (52.9–67.7)

Australasian taxa (node 15) 59.2 (54.4–64.0) 62.2 (54.7–69.3) 61.5 (54.9–67.9) 62.8 (56.9–70.6)

Australidelphia (node 16) 59.9 (55.0–64.8) 62.9 (55.3–70.0) 61.5 (54.9–67.8) 62.8 (56.9–70.6)

Paucituberculata (node 17) 9.2 (6.9–11.8) 9.7 (7.1–12.7) 13.9 (5.5–24.4) 14.1 (4.9–25.5)

All marsupials but Didelphimorphia(node 18)

71.3 (64.6–78.5) 74.8 (65.1–84.6) 83.3 (68.8–100.7) 85.6 (69.5–104.4)

Didelphini + Marmosini (node 19) 26.6 (21.3–32.5) 27.9 (21.8–34.8) 25.4 (13.4–41.4) 25.0 (13.5–39.9)

Didelphimorphia (node 20) 35.8 (29.6–42.5) 37.5 (30.4–45.6) 35.7 (21.1–52.1) 35.4 (21.5–51.6)

Marsupialia (node 21) 74.8 (67.0–83.0) 78.5 (67.8–89.4) 86.8 (71.8–103.8) 89.3 (72.1–108.3)

a Values are in millions of yearsb Node numbers refer to Fig. 2c Represented by Petauridae and Pseudocheiridae

J Mammal Evol

Chiroptera fossil records, the marsupial UBBL record is relatively more complete. Theaustralidelphian UBBL is approximately 5% (BEAST) to 9% (Multidivtime) less complete thanthe ameridelphian fossil record. Of the seven marsupial orders, the notoryctemorphian UBBLis the most incomplete with over 67% missing and the Microbiotheria record is the mostcomplete with no more than 5% missing. Among Australasian orders, the record forDiprotodontia is the most complete (42–43% missing).

Discussion

Phylogenetic relationships

Marsupial cohorts and the root of Marsupialia

The seven marsupial orders are currently divided into the cohorts Ameridelphia andAustralidelphia (Szalay 1982). We found strong support for the monophyly of Australidelphiaand its constituent orders, even with the addition of more australidelphian taxa to the Amrine-Madsen et al. (2003) data set. Australidelphian monophyly is in solid agreement with previousmorphological (Luckett 1994; Szalay and Sargis 2001; Horovitz and Sánchez-Villagra 2003),molecular (Kirsch et al. 1991, 1997; Springer et al. 1998; Phillips et al. 2001, 2006; Amrine-Madsen et al. 2003), and mixed data sets (Asher et al. 2004). Hershkovitz (1992) suggestedthat the continuous lower ankle joint pattern used by Szalay (1982) to define Australidelphia isnot a synapomorphy because it is present in some ameridelphian taxa. Szalay (1994) extended

Multidivtime Partitioned

y = 0.2725x

R2 = 0.7805

0

5

10

15

20

25

0 20 40 60 80

Mean Time (MA)

95%

Con

fide

nce

Inte

rval

Wid

th (

MA

)

Fig. 3 Ninety-five percent confidence intervals plotted against mean for divergence date estimates for hardbounded Multidivtime analysis (diprotodontian maximum=65Ma). Raw data are given in the SupplementaryInformation.

J Mammal Evol

his pedal character analyses and demonstrated that virtually all australidelphians display thisankle joint pattern. Luckett (1994) agreed with Szalay (1982, 1994) and suggested that adouble or triple faceted calcaneocuboid may also support Australidelphia. The continuouslower ankle joint pattern (Szalay 1982, 1994) and the 12 unambiguous cranial and postcranialcharacters listed by Horovitz and Sánchez-Villagra (2003) appear to be reliable morphologicalcharacters that define Australidelphia.

Sperm pairing is often cited in support of Ameridelphia, but in didelphimorphians it is side-by-side whilst in the paucituberculates it is head-to-head (Temple-Smith 1987). Themorphological study of Horovitz and Sánchez-Villagra (2003) found no characters in supportof Ameridelphia. Our results provide little or no support for Ameridelphia and suggest that theroot of the marsupial tree is between Didelphimorphia and Paucituberculata + Australidelphia(ML, Bayesian) or between Paucituberculata and Didelphimorphia + Australidelphia (MP). Asister-group relationship between paucituberculates and australidelphians to the exclusion of

Table 6 Estimation of the missing fossil record for marsupial, felid, and chiropteran lineages

Percent missinga Average of percentage missing per lineage

Marsupialiab

All lineages 45.4/48.0 65.8/67.0

External lineages 45.5/46.9 50.5/52.7

Internal lineages 45.1/49.5 82.7/82.7

Dasyuridae 50.8/58.4 62.1/64.8

Diprotodontia 42.6/41.7 59.3/57.5

Notoryctemorphia 67.0/67.5 67.0/67.5

Peramelidae 57.1/60.1 61.4/70.2

Microbiotheria 4.9/4. 8 4.9/4. 8

Paucituberculata 29.2/40.0 69.4/72.1

Didelphimorphia 51.6/52.6 70.7/71.6

All Australidelphia but Diprotodontia 46.4/50.0 65.5/69.3

American taxa 39.8/44.6 69.7/70.8

Australasian taxa 48.7/50.2 63.3/64.5

Australidelphia 44.4/45.7 62.6/63.7

Ameridelphia 47.6/52.6 76.2/77.5

Felidaec

All lineages 70 76

External lineages 70 72

Internal lineages 69 80

Chiropterad

All lineages 61 73

External lineages 59 58

Internal lineages 71 89

a Summed over all lineagesb Values for marsupial lineages are from Multidivtime and BEAST partitioned analyses, respectively, with thediprotodontian maximum set to 65.0Ma. Raw data for marsupial lineages are given in Table 5.c Values calculated from those given in Johnson et al. (2006)d Values from Teeling et al. (2005)

J Mammal Evol

didelphimorphians is in agreement with results obtained using complete mitochondrialgenomes (Nilsson et al. 2003, 2004), concatenated nuclear genes with fewer taxa (Amrine-Madsen et al. 2003), morphological data (Horovitz and Sánchez-Villagra 2003), and mixeddata sets (Asher et al. 2004). Horovitz and Sánchez-Villagra (2003) cited eight cranial andskeletal characters in support of Paucituberculata + Australidelphia. Robust resolution of theroot of the marsupial tree will likely await the discovery of rare genomic changes definingparticular groups (e.g., indels).

Didelphimorphia

Our analysis included only three didelphimorphian lineages. We recovered Caluromys as thesister taxon to a Marmosinae + Didelphinae clade. This result is consistent with previousmorphological and molecular studies (Reig et al. 1987; Kirsch et al. 1997; Steiner et al. 2005;Jansa and Voss 2005).

Australidelphia

Delineation of the basal split within Australidelphia has proven difficult to resolve. Severalmolecular and morphological studies have recovered microbiotheres as nested somewherewithin the Australasian radiation (e.g., Kirsch et al. 1991, 1997; Burk et al. 1999; Szalay andSargis 2001; Horovitz and Sánchez-Villagra 2003; Nilsson et al. 2003, 2004; Munemasa et al.2006). This is in marked contrast to our nuclear gene results and to the most recent molecularstudies (e.g., Amrine-Madsen et al. 2003; Phillips et al. 2006), which consistently recovermicrobiotheres as the sister-taxon to a monophyletic Australasian clade. This suggests that theAustralasian clade can be defined by the presence of an epitympanic sinus in the squamosalabove the auditory meatus (Woodburne 1984) and that the 15 supposedly unambiguousapomorphies uniting Dromiciops and Diprotodontia given by Horovitz and Sánchez-Villagra(2003) are either convergent and/or plesiomorphic for Australidelphia. Furthermore, our resultshave significance for understanding the biogeographic history of marsupials and suggest theneed for only a single dispersal event from South America to Australia, presumably throughAntarctica. A nested position of Dromiciops with Australidelphia would suggest multipledispersion events and/or back migrations.

Within the Australasian clade, we find robust support for all recognized orders. There is alsosupport for a basal split between Diprotodontia and a clade consisting of Peramelemorphiatogether with Notoryctemorphia and Dasyuromorphia. This result is consistent with theAmrine-Madsen et al. (2003) nuclear analysis and with the combined nuclear plusmitochondrial and mitochondrial only DNA analyses of Phillips et al. (2006).

Dasyuromorphia and Peramelidae

Within Dasyuromorphia we find strong support for Phascogalini (Phascogale + Antechinus), asister-group relationship between Phascogalini and Dasyurini (Dasyurus), and a basal splitbetween Sminthopsinae and the Dasyurini + Phascogalini clade. These results are consistentwith those of nuclear, mitochondrial, and DNA hybridization studies (Kirsch et al. 1990, 1997;Krajewski et al. 2000). In contrast Wroe et al. (2000) recovered a basal split between theDasyurini and the other dasyurids and Phascogalini paraphyly. Within the Peramelidae the splitbetween Peramelinae (Perameles + Isoodon) and Echymiperinae (Echymipera) is robustlysupported. This is consistent with mitochondrial (Westerman et al. 1999, 2001) and DNAhybridization studies (Kirsch et al. 1997).

J Mammal Evol

Diprotodontia

Historically, relationships between sections of Diprotodontia, the largest and most diverse orderof Australasian marsupials, have been difficult to resolve. Members of this order are united byseveral distinctive synapomorphies: diprotodonty, a superficial thymus, syndactyly, a fasciculusaberrans connecting the two hemispheres of the brain (Abbie 1937), and 22 additionalmorphological apomorphies given by Horovitz and Sánchez-Villagra (2003). Despite theoverwhelming morphological evidence in support of this clade, robust molecular support hasonly recently been found (Amrine-Madsen et al. 2003). Our study adds one diprotodontianfamily (Burramyidae) to the Amrine-Madsen et al. (2003) data set and finds robust support forDiprotodontia.

The interrelationships of diprotodontian families and superfamilies remain controversial.Kirsch et al. (1997) and Wilson and Reeder (2005) recognized three suborders within theDiprotodontia: Macropodiformes (kangaroos and kin), Phalangeriformes (possums and kin),and Vombatiformes (wombats and koalas). Our results show robust support for Vombatiformes(Vombatus + Phascolarctos). Molecular support for the monophyly of Vombatiformes isconsistent with their hook-shaped spermatozoa (Hughes 1965; Harding 1987), serological data(Kirsch 1968, 1977), DNA hybridization (Springer and Kirsch 1991; Kirsch et al. 1997;Springer et al. 1997a, b), mitochondrial DNA (Burk et al. 1999; Kavanagh et al. 2004;Munemasa et al. 2006), nuclear DNA (Amrine-Madsen et al. 2003), morphology (Horovitz andSánchez-Villagra 2003), and combined molecular and morphological studies (Asher et al.2004). We also find robust molecular support for Macropodiformes + Phalangeriformes(= Phalangerida of Aplin and Archer 1987), which is consistent with some morphologicalevidence, including posterior expansion of the alisphenoid tympanic wing (Winge 1941;Springer and Woodburne 1989), and a recent analysis of mitochondrial genome sequences(Munemasa et al. 2006).

Whereas the monophyly of Macropodiformes was strongly supported, we recovered nosupport for the monophyly of Australasian possums (i.e., Phalangeriformes). In contrast toDNA hybridization studies (Springer and Kirsch 1991; Kirsch et al. 1997), which supportedAustralasian possum monophyly (i.e., Phalangeriformes), our results argue against themonophyly of this clade. The lack of molecular support for Phalangeriformes suggests thatmorphological characters supporting this clade, such as a tube-like ectotympanic that is fusedto other bones of the skull (Flannery 1987; Springer and Woodburne 1989), are convergent inphalangeroids and petauroids.

The association of pseudocheirids and petaurids (Petauroidea) has been recovered withmolecular data sets (e.g., Kirsch 1977; Kirsch et al. 1997; Osborne and Christidis 2001;Osborne et al. 2002; Amrine-Madsen et al. 2003; Kavanagh et al. 2004), mixed data sets(Asher et al. 2004), and some morphological data sets (e.g., Archer et al. 1999). Kirsch et al.(1997) considered petaurids and pseudocheirids among the most closely related diprotodontianfamilies. However, the morphological analysis of Horovitz and Sánchez-Villagra (2003)recovered Petaurus as the sister taxon to all other diprotodontians and Woodburne et al. (1987)suggested the pseudocheirids were the sister group to all other possums. All of ourphylogenetic and statistical analyses unambiguously supported Petauroidea, albeit with limitedtaxon representation.

Determination of the phylogenetic affinities of the burramyids has been less forthcoming.Molecular studies have recovered them as the sister group to the Vombatiformes (Osborne et al.2002), sister to all other possums (Edwards and Westerman 1995), or sister to the phalangerids(Springer and Kirsch 1991; Kirsch et al. 1997). Gunson et al. (1968) suggested an associationwith Acrobatidae based on chromosome number. Morphological studies have likewise been

J Mammal Evol

ambiguous. Springer and Woodburne (1989) and Marshall et al. (1990) suggested anassociation with the petauroids but pedal morphology suggests an association with theacrobatids (Szalay 1994). In the absence of acrobatid exemplars, our analyses suggest that theburramyids are the sister group to the phalangerids, although statistical tests could notdiscriminate between different phylogenetic hypotheses.

Timeline for marsupial evolution

BEAST and multidivtime comparisons

For groups with a poor fossil record (e.g., Australasian marsupials), molecular estimates ofdivergence times are of critical importance in helping to understand the evolution, timing, andpossible causes of radiation. The present study is the first to employ two different relaxedmolecular clock methods to a marsupial nuclear gene data set that includes representatives ofall marsupial orders in addition to multiple placental outgroups. Both of the Bayesian datingmethods (BEAST, Multidivtime) employ a relaxed molecular clock assumption and allow theincorporation of multiple fossil constraints. The methods differ in that BEAST simultaneouslyestimates the phylogeny and divergence dates, and allows for soft-bounded constraints oncalibration nodes, whereas Multidivtime requires an optimal tree to be specified and onlypermits hard-bounded constraints. Even with these differences, there is still good agreementbetween most divergence dates that were estimated by these methods.

Soft-bounded node constraints have the advantage of allowing sequence data to correct poornode constraints if they exist. We obtained similar divergence time estimates with soft and hardbounds in BEAST analyses, which suggest that fossils are not in conflict with either themselvesor the molecular data (Yang and Rannala 2006).

Drummond et al. (2006) analyzed the Amrine-Madsen et al. (2003) marsupial data set. TheirMAP topology obtained from a non-partitioned UCLN analysis is different from ours in thatNotoryctes is the sister taxon to a Peramelidae + Dasyuromorphia clade, Dromiciops is thesister taxon to the Diprotodontia, and within Diprotodontia the petauroids are the sister groupto the kangaroos. For all of the nodes that are directly comparable to our phylogeny (Fig. 1),the estimated divergence dates are similar (Table 7). Drummond et al. (2006) found that thefastest branch on their tree was evolving 2.7× faster than the slowest branch. Our results for allof the non-partitioned analyses indicate that the branch leading to the Dasyuromorphia (L22) isevolving 2.8× faster than the slowest branch (L15, branch leading to Diprotodontia). In theMultidivtime non-partitioned analyses, the Dasyurus rate (L19) was 4.7× faster than theCercartetus rate (L4).

Nilsson et al. (2003, 2004) reconstructed phylogenetic relationships among marsupial ordersusing complete mitochondrial genomes and also inferred a timeline for marsupial evolutionwith relaxed clock dating methods. The phylogeny of Nilsson et al. (2003, 2004) differs fromour Bayesian tree (Fig. 1) by supporting Marsupionta (monotremes plus marsupials),Dromiciops gliroides as the sister taxon to a Peramelemorphia + Dasyuromorphia +Notoryctemorphia clade, and Vombatiformes as the sister taxon to Macropodiformes. Nilssonet al. (2004) also used a maximum constraint of 135 Ma for the base of Mammalia, which ismuch younger than dates suggested by other molecular studies that do not constrain this node(Woodburne et al. 2003). Given these differences, it is not surprising that our estimated datesare generally older than the dates of Nilsson et al. (2003, 2004) (Table 7). Nilsson et al. (2004)estimate the root of Marsupialia at 69 Ma, which is younger than our estimates byapproximately 6–20 Ma. They also suggest an age of only 42 Ma for the base of the

J Mammal Evol

Peramelemorphia + Dasyuromorphia + Notoryctemorphia clade, whereas our estimate for thissplit is 57–62 Ma. The date suggested by our analysis is consistent with putative stemdasyuromorphian incertae sedis fossils and the “perameloid” tooth known from 54.6 Ma oldMurgon deposits (Godthelp et al. 1999). Nilsson et al. (2003, 2004) also suggest thataustralidelphians last shared a common ancestor 50–51 Ma. This age is younger than themicrobiothere fossil record, which extends as far back as the middle Paleocene (e.g., Khasia).Even if one does not accept Khasia as a microbiothere, definitive australidelphians fromMurgon at 54.6 Ma (Godthelp et al. 1999) indicate an older age for Australidelphia thansuggested by Nilsson et al. (2003, 2004).

All of our Bayesian estimates suggest that crown group marsupials had a most recentcommon ancestor in the Cretaceous, possibly as far back as 89 Ma. These estimates areyounger than Hasegawa et al.’s (2003) estimate of ∼100 Ma. Benton and Donoghue (2007)provide 30 hard minima and soft maxima fossil constraints for dating the tree of life. One oftheir proposed dates is the opossum–kangaroo split, i.e., the ameridelphian–australidelphiansplit. They suggest a hard minimum of 61.5 Ma and a soft maximum of 71.2 Ma. They basethe minimum on the newly described probable Paleocene polydolopimorph, Cocatherium(Goin et al. 2006), and claim this is the oldest known crown-group marsupial. However, ifthese authors accept that Cocatherium is a polydolopimorph, and that polydolopimorphs arecrown-group metatherians, then this cannot be the oldest member of crown-group Marsupialiaas the polydolopimorph fossil record extends back to the Judithian (74–79 Ma; Case et al.2005). In contrast, Goin et al. (2006) recovered Polydolopimorphia as stem metatherians ratherthan as members of the crown-group. Case et al. (2005) found that Nortedelphys (Lancian) wasthe oldest crown-group metatherian. Our estimated dates for the base of Marsupialia areconsistent with younger fossil dates given that the fossil record usually underestimates the truetime of divergence for a given node.

Our molecular data suggest an Eocene date for the origin of the living didelphimorphians.Our estimates are slightly younger than those obtained by Steiner et al. (2005) using a singlenuclear gene (transthyretin), and approximately 14 Ma younger than complete mitochondrialgenome estimates (Nilsson et al. 2004; Table 7). Steiner et al. (2005) attributed the middleEocene origin of living didelphids to one of the first definitive phases of the Andean uplift,which caused a cooling and drying out of the woodland habitats. Our divergence time estimatefor the Didelphinae + Marmosinae split is consistent with this interpretation.

A secure phylogeny for Australidelphia is of paramount importance for understanding theearly biogeographic history of this group. As previously mentioned, some molecular andmorphological studies recover microbiotheres as nested somewhere within the Australasianradiation (e.g., Kirsch et al. 1991, 1997; Nilsson et al. 2003, 2004; Horovitz and Sánchez-Villagra 2003). The nesting of Microbiotheria within the Australasian marsupial radiation is inmarked contrast to our results above and to other recent molecular studies (e.g., Amrine-Madsen et al. 2003; Phillips et al. 2006) that recover microbiotheres as the sister-taxon to amonophyletic Australasian clade. A sister relationship between microbiotheres and Austral-asian marsupials, together with our estimated dates of divergence (60–65 Ma) for this split,suggest that the paleobiogeography of Australidelphia involves a single dispersal event fromSouth America to Australia via Antarctica. Monophyly of the Australasian marsupials makesfor a much simpler biogeographic history and may explain the complete absence ofAustralasian taxa from the middle Eocene deposits of Antarctica––despite the postulatedimportance of Antarctica as a venue for the early diversification of Gondwanan marsupials.Furthermore, the necessity of a pan-Gondwanan distribution of marsupials before theseparation of Antarctica and Australia with subsequent vicariance and extinction is notrequired. The most recent cladistic analysis of early metatherians has recovered the South

J Mammal Evol

American species Chulpasia mattaueri as a stem australidelphian (Goin et al. 2006) and thereappears to be an undescribed species from the Murgon deposit (54.6 Ma) in Australia (Sigé etal. 1995). Chulpasia mattaueri is known from the Umayo Formation, which according to Sigéet al. (1995) is most consistent in age with chron 24 (latest Paleocene or earliest Eocene).These fossils are slightly younger than our expected date for stem australidelphians, butnonetheless are consistent with our study in that we would expect the fossil record tounderestimate the true times of divergence.

Australasian marsupials

As currently understood the Australian plate migrated northward after its complete separationfrom Antarctica and by the middle Oligocene had collided with the Asian plate in the NewGuinean region resulting in the emergence of New Guinea and the uplift of the New GuineanHighlands. By the late Oligocene the Sepik Province had docked with the emerging land mass.The rising of the New Guinean Central Cordillera along with the emergence of Timor and theestablishment of the circumpolar current to the south of Australia had a major effect onclimates across Australia and changed the drainage patterns of central Australia leading to aprogressive drying. In Australia, the average temperature was lowered and the rainforestsbegan to be replaced by drier forest types and more open country (White 1994; Heads 2002).By the late Miocene, xeric and mesic habitats predominated much of central Australia. By thelate Pliocene there were periods of prolonged cooling.

Table 7 Molecular divergence dates from other relaxed clock studies

Nilsson et al. 2003a Nilsson et al. 2004b Drummond et al. 2006c

Node

1. Macropodidae + Potoroidae N/A 20 N/A

4. Vombatiformes N/A N/A 33

5. Macropodiformes + Phalangeroidea 33 N/A N/A

6. Macropodiformes + Phalangeriformes N/A N/A 41

7. Diprotodontia 42 46 48

9. Dasyurini + Phascogalini N/A N/A 15

10. Dasyuridae N/A 20 25

11. Notoryctes + Dasyuridae N/A 39 N/A

13. Peramelidae N/A 11 16

14. Peramelidae + Notoryctes + Dasyuridae N/A 42 62

16. Australidelphia 51 51 66

17. Paucituberculata 29 31 16

18. All marsupials but Didelphimorphia 62 61 N/A

19. Didelphinae+ Marmosinae 52 52 29

20. Didelphimorphia NA NA 39

21. Marsupialia 64 69 85

Dates are in millions of yearsa Dates are estimated from their Fig. 3b Dates are estimated from their Fig. 3c Nodes numbers correspond to our Fig. 1. Nodes not shown were not present in the data sets listed above.

J Mammal Evol

Case (1989) suggested that all of the major lineages within the Diprotodontia were presentin the Eocene. In the late Eocene the dominant podocarp forests began to be replaced byNothofagus dominated forests, which “opened” up the forest canopy. This floristic change mayhave promoted the radiation of the arboreal possum species in the middle to late Eocene.Radiation of the terrestrial forms probably did not occur until the forests began to open up withan herbaceous angiosperm understorey in the Oligocene. Megirian et al. (2004) and Adam(1999) provide evidence that rainforests were almost certainly not the predominant vegetationin the North Territory, South Australia, and NW Queensland from the late Oligocene onward. Itis then probable that the terrestrial australidelphians began to radiate before the late Oligocene.As the grasslands spread concurrently with the “drying out” of Australia in the late Mioceneand early Pliocene, more specialized herbivores such as macropodine grazers evolved tooccupy new ecological niches. Our estimated dates of divergence are congruent with suchscenarios. The terrestrial dasyuromorphians and peramelemorphians radiate in the lateOligocene and early Miocene. All of the major diprotodontian lineages in our study werepresent in the early Tertiary. The clades of arboreal possums (Petauroidea and Phalangeroidea)emerged before the late Oligocene/early Miocene. The split between the terrestrial/fossorialvombatid ancestor and the arboreal phascolarctids again occurred before the canopy began toopen up. Furthermore, the split between the potoroids (non-grazing kangaroos) and thepredominantly grazing kangaroos occurred ∼14–17 Ma, later than any of the other interfamilialdivergences within Diprotodontia.

Unrepresented basal branch lengths

The UBBL technique for estimating the completeness of the fossil record has previously beenapplied to two mammalian clades—Chiroptera (Teeling et al. 2005) and Felidae (Johnson et al.2006). A similar method has also been applied to the Echinoidea (Smith et al. 2006). Table 6shows that the percentage of the UBBL that is missing, as well as the average percentagemissing per lineage, is higher for Felidae and Chiroptera than for Marsupialia. Felids arecarnivores and marsupials are predominantly herbivores and omnivores. In general, herbivoreshave a greater biomass (being lower on the food chain) and thus have a correspondingly higherchance of fossilization. In the case of bats, many taxa live in tropical environments wherepreservation rates are low. Small body size and fragile bones also help to explain the poorfossil record of bats.

Table 6 shows that the unrepresented basal branch length is more incomplete orapproximately equivalent for both internal and external lineages in all three clades. Thisfinding is consistent with the general rule that the fossil record becomes more incompletefurther back in time. The lack of diagnostic characters for internal branches, which representdirect ancestors and/or extinct side branches, may also contribute to the greater unrepresentedbasal branch length for internal branches than for external branches.

Among marsupials, it is perhaps surprising that our estimates suggest that the Australasianmarsupial record is approximately as complete as the South American record even though theliterature is filled with claims that underscore the incompleteness of the Australasian fossilrecord. However, marsupials have inhabited South America for a longer period of time and, asnoted above, the completeness of the fossil record decreases with increasing time spans. Also,estimates for the incompleteness of the South American record may diminish pending thephylogenetic placement of key fossil taxa. We recovered a date of 35–39 Ma for the didelphid–caluromyid split, which resulted in an estimate of 100% incompleteness for the long branch(branch L38, Fig. 2) leading to Caluromys. However, some authors (Reig et al. 1987;McKenna and Bell 1997) have suggested that the early Miocene genus Pachybiotherium

J Mammal Evol

belongs to Caluromyinae, which would reduce the incompleteness of this lineage byapproximately 50%.

Our estimates for the completeness of the South American record assume that several LateCretaceous and early Paleocene fossils are crown group marsupials (Goin et al. 2006). IfRougier et al. (1998), Luo et al. (2003), Horovitz and Sánchez-Villagra (2003), and Sánchez-Villagra et al. (2007) are correct in their contention that there are no known crown groupmarsupials from the Late Cretaceous or early Paleocene, then the completeness of the recordfor South American orders becomes even more impoverished.

Among the South American marsupial orders, the apparent completeness of the micro-biothere record is surprising if the single living species (Dromiciops gliroides) is taken as ananalog for all microbiotheres. D. gliroides has a restricted geographic distribution and lives inan environment that is not conducive to fossilization. The completeness of the microbiotherefossil record is based on the genus Khasia, which is known from 60.4 to 59.2 Ma old Boliviandeposits. Although Khasia is believed to be a microbiothere (e.g., Woodburne and Case 1996;Marshall et al. 1997), Wroe et al. (2000) have called the identification of fossil microbiotheresinto question. These authors argue that the characters used to diagnose microbiotheres areessentially symplesiomorphic or convergent and that microbiotheres are not well diagnoseddentally. If true, this would complicate the correct identification of fossil species such asKhasia that are known only from dental elements. However, Wroe et al. (2000) allow for thepossibility that these fossils are microbiotheres.

Acknowledgements We dedicate this paper to the memory of our friend and colleague John A. W. Kirsch. MSSacknowledges support from NSF.

References

Abbie AA (1937) Some observations on the major subdivisions of the Marsupialia, with special reference to theposition of the Peramelidae and Caenolestidae. J Anat 71:424–436

Adam P (1999) Australian rainforests. Oxford University Press, OxfordAmrine-Madsen H, Scally M, Westerman M, Stanhope MJ, Krajewski C, Springer MS (2003) Nuclear gene

sequences provide evidence for the monophyly of australidelphian marsupials. Mol Phylogenet Evol28:186–196

Aplin KP, Archer M (1987) Recent advances in marsupial systematics with a new syncretic classification. In:Archer M (ed) Possums and opossums: studies in evolution. Surrey Beatty and Sons, Chipping Norton, NewSouth Wales, pp xv–xxii

Archer M (1982) Review of the dasyurid (Marsupialia) fossil record, integration of data bearing on phylogeneticinterpretation, and suprageneric classification. In: Archer M (ed) Carnivorous marsupials. Royal ZoologicalSociety of New South Wales, Mossman, New South Wales, pp. 397–443

Archer M, Hand SJ (2006) The Australian marsupial radiation. In: Merrick JR, Archer M, Hickey GM, LeeMSY (eds) Evolution and biogeography of Australasian vertebrates. Auscipub, New South Wales, pp575–646

Archer M, Kirsch JAW (2006) The evolution and classification of marsupials. In: Armati PJ, Dickman CR,Hume ID (eds) Marsupials. Cambridge University Press, New York, pp 1–21

Archer M, Wade M (1976) Results of the Ray E Lemley Expeditions: part 1 the Allingham Formation and a newPliocene vertebrate fauna from northern Queensland Australia. Mem Qld Mus 17:379–398

Archer M, Tedford RH, Rich TH (1987) The Pilkipildridae, a new family and new species of? petauroid possums(Marsupialia: Phalangerida) from the Australian Miocene. In: Archer M (ed) Possums and opossums: studiesin evolution. Surrey Beatty and Sons, Chipping Norton, New South Wales, pp 607–628

Archer M, Arena R, Bassarova M, Black K, Brammall J, Cooke B, Creaser P, Crosby K, Gillespie A, GodthelpH, Gott M, Hand SJ, Kear B, Krikmann A, Mackness B, Muirhead J, Musser A, Myers T, Pledge N, WangY, Wroe S (1999) The evolutionary history and diversity of Australian mammals. Aust Mammal 21:1–45

Asher RJ, Horovitz I, Sánchez-Villagra MR (2004) First combined cladistic analysis of marsupial mammalinterrelationships. Mol Phylogenet Evol 33:240–250

J Mammal Evol

Benton MJ, Ayala FJ (2003) Dating the tree of life. Science 300:1698–1700Benton MJ, Donoghue PCJ (2007) Paleontological evidence to date the tree of life. Mol Biol Evol 24:26–53Brammall JR (1998) A new petauroid possum from the Oligo-Miocene of Riversleigh, northwestern Queensland.

Alcheringa 23:31–50Brammall J, Archer M (1999) Living and extinct petaurids, acrobatids, tarsipedids and burramyids (Marsupialia):

relationships and diversity through time. Aust Mammal 21:24–25BurkA,WestermanM,KaoDJ, Kavanagh JR, SpringerMS (1999)An analysis ofmarsupial interordinal relationships

based on 12S rRNA, tRNAvaline, 16S rRNA, and cytochrome b sequences. J Mammal Evol 6:317–334Case JA (1984) A new genus of Potoroinae (Marsupialia Macropodidae) from the Miocene Ngapakaldi Local

Fauna, South Australia, and a definition of the Potoroinae. J Paleontol 58:1074–1086Case JA (1989) Antarctica: The effect of high latitude heterochroneity on the origin of the Australian marsupials. In:

Crame JA (ed) Origins and evolution of the Antarctic biota. Geological Society Special Publication 47, pp 217–226Case JA (2001) Turnover of bandicoots in the Oligo-Miocene of South Australia. J Vertebr Paleontol Supp

21:39ACase JA, Goin FJ, Woodburne MO (2005) “South American” marsupials from the Late Cretaceous of North

America and the origin of marsupial cohorts. J Mammal Evol 12:461–494Cooke BN (2006) Kangaroos. In: Merrick JR, Archer M, Hickey GM, Lee MSY (eds) Evolution and

biogeography of Australasian vertebrates. Auscipub, New South Wales, pp 647–672Cooke B, Kear B (1999) Evolution and diversity of kangaroos (Macropodoidea, Marsupialia). Aust Mammal

21:27–29Cozzuol MA, Goin F, Reyes MDL, Ranzi A (2006) The oldest species of Didelphis (Mammalia, Marsupialia,

Didelphidae), from the late Miocene of Amazonia. J Mammal 87:663–667Crochet J-Y, Sigé B (1993) Les mammifères de Chulpas (Formation Umayo, transition Crétacé-Tertiaire, Pérou):

données préliminaires. Documents Lab Géol Faculté Sci Lyon 125:97–107Crosby K, Bassarova M, Archer M, Carbery K (2004) Fossil possums in Australasia: discovery, diversity and

evolution. In: Goldingay RL, Jackson SM (ed) The biology of Australian possums and gliders. Beatty,Sydney, pp 161–176

Czaplewski NJ, Masanaru T, Naeher, TM, Shigehara N, Setoguchi T (2003) Additional bats from the middleMiocene La Venta fauna of Colombia. Rev Acad Colomb Cienc 27:263–282

De Queiroz A (1993) For consensus (sometimes). Syst Biol 42:368–372Drummond AJ, Rambaut A (2003) BEAST version 14 [computer program] Available http://evolvezoooxacuk/

beast Accessed January 2007Drummond AJ, Ho SYW, Phillips MJ, Rambaut A (2006) Relaxed phylogenetics and dating with confidence.

PLoS Biol 4:699–710Edwards D, Westerman M (1995) The molecular relationships of possum and glider families as revealed by

DNA–DNA hybridisations. Aust J Zool 43:231–240Farris JS, Kallersjo M, Kluge AG, Bult C (1994) Testing significance of incongruence. Cladistics 10:315–319Flannery T (1987) The relationships of the macropodoids (Marsupialia) and the polarity of some morphological

features within the Phalangeriformes. In: Archer M (ed) Possums and opossums: studies in evolution. SurreyBeatty and Sons, Chipping Norton, New South Wales, pp 741–748

Flannery TF (1989) Phylogeny of the Macropodoidea; a case study in convergence. In: Grigg GC, Jarman PJ,Hume ID (eds) Kangaroos, wallabies and rat–kangaroos. Surrey Beatty and Sons, Chipping Norton, NewSouth Wales, pp 1–46

Godthelp H, Wroe S, Archer M (1999) A new marsupial from the early Eocene Tingamarra local Fauna ofMurgon, southeastern Queensland: a prototypical Australian marsupial? J Mammal Evol 6:289–313

Goin FJ (1997) New clues for understanding Neogene marsupial radiations. In: Kay RF (ed) Vertebratepaleontology in the Neotropics: the Miocene fauna of La Venta, Colombia. Smithsonian Institution,Washington, DC, pp 187–206

Goin FJ, Pascual R, Tejedor MF, Gelfo JN, Woodburne MO, Case JA, Reguero MA, Bond M, Lopez GM, CioneAL, Udrizar Sauthier D, Balarino L, Scasso RA, Medina FA, Ubaldon MC (2006) The earliest Tertiarytherian mammal from South America. J Vertebr Paleontol 26:505–510

Gradstein FM, Ogg JG, Smith AG, 39 co-authors (2004) A geologic timescale. Cambridge University Press,Cambridge

Guindon S, Gascuel O (2003) A simple, fast, and accurate algorithm to estimate large phylogenies by maximumlikelihood. Syst Biol 52:696–704

Gunson MM, Sharman GB, Thomson JA (1968) The affinities of Burramys (Marsupialia: Phalangeroidea) asrevealed by its chromosomes. Aust J Sci 31:40–41

Harding HR (1987) Interrelationships of the families of the Diprotodontia—a view based on spermatozoanultrastructure. In: Archer M (ed) Possums and opossums: studies in evolution. Surrey Beatty and Sons,Chipping Norton, New South Wales, pp 195–216

J Mammal Evol

Hasegawa M, Thorne JL, Kishino H (2003) Time scale of eutherian evolution estimated without assuming aconstant rate of molecular evolution. Genes Genet Syst 78:267–283

Heads M (2002) Birds of paradise, vicariance biogeography and terrane tectonics in New Guinea. J Biogeog29:261–283

Hedges SB, Kumar S (2004) Precision of molecular time estimates. Trends Gen 20:242–247Hershkovitz P (1992) Ankle bones: the Chilean opossum Dromiciops gliroides Thomas, and marsupial

phylogeny. Bonn Zool Beit 43:181–213Horovitz I, Sánchez-Villagra MR (2003) A morphological analysis of marsupial mammal higher-level

phylogenetic relationships. Cladistics 19:181–212Huelsenbeck JP, Ronquist F (2001) MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics

17:754–755Hughes RL (1965) Comparative morphology of spermatozoa from five families. Aust J Zool 13:533–543Jansa SA, Voss RS (2005) Phylogenetic relationships of the marsupial genus Hyladelphys based on nuclear gene

sequences and morphology. J Mammal 86:853–865Johnson WE, Eizirik E, Pecon-Slattery J, Murphy WJ, Antunes A, Teeling E, O’Brien SJ (2006) The late

Miocene radiation of modern Felidae: a genetic assessment. Science 311:73–77Kavanagh JR, Burk-Herrick A, Westerman M, Springer MS (2004) Relationships among families of

Diprotodontia (Marsupialia) and the phylogenetic position of the autapomorphic honey possum (Tarsipesrostratus). J Mammal Evol 11:207–222

Kielan-Jaworowska Z, Cifelli RL, Luo Z-X (2004) Mammals from the age of dinosaurs: origins, evolution, andstructure. Columbia University Press, New York

Kirsch JA (1968) Prodromus of the comparative serology of Marsupialia. Nature 217:418–420Kirsch JAW (1977) The comparative serology of Marsupialia and a classification of marsupials. Aust J Zool,

Supp Ser 52:1–152Kirsch JAW, Palma RE (1995) DNA/DNA hybridization studies of carnivorous marsupials. V. A further estimate

of relationships among opossums (Marsupialia: Didelphidae). Mammalia 59:403–425Kirsch JAW, Krajewski C, Springer MS, Archer M (1990) DNA–DNA hybridization studies of carnivorous

marsupials. II. Relationships among dasyurids (Marsupialia: Dasyuridae). Aust J Zool 38:673–696Kirsch JAW, Dickerman AW, Reig OA, Springer MS (1991) DNA hybridization evidence for the Australasian

affinity of the American marsupial Dromiciops australis. Proc Natl Acad Sci USA 88:10465–10469Kirsch JAW, Lapointe F-J, Springer MS (1997) DNA-hybridisation studies of marsupials and their implications

for metatherian classification. Aust J Zool 45:211–280Kishino H, Hasegawa M (1989) Evaluation of the maximum likelihood estimate of the evolutionary tree

topologies from DNA sequence data and the branching order in Hominoidea. J Mol Evol 29:170–179Kishino H, Thorne JL, Bruno WJ (2001) Performance of a divergence time estimation method under a

probabilistic model of rate evolution. Mol Biol Evol 18:352–361Krajewski C, Young J, Buckley L, Woolley PA, Westerman M (1997) Reconstructing the evolutionary radiation of

dasyurine marsupials with cytochrome b, 12S rRNA, and protamine P1 gene trees. J Mammal Evol 4:217–236Krajewski C, Blacket MJ, Westerman M (2000) DNA sequence analysis of familial relationships among

dasyuromorphian marsupials. J Mammal Evol 7:95–108Kumar S, Hedges SB (1998) A molecular timescale for vertebrate evolution. Nature 392:917–920Langford RP, Wilford GE, Truswell EM, Isern AR (1995) Palaeogeographic atlas of Australia, Volume 10

Cainozoic. Australian Geological Survey Organization, CanberraLong J, Archer M, Flannery TF, Hand SJ (2002) Prehistoric mammals of Australia and New Guinea: one

hundred million years of evolution. University of New South Wales Press, SydneyLuckett WP (1994) Suprafamilial relationships within Marsupialia: resolution and discordance from

multidisciplinary data. J Mammal Evol 2:255–283Luo Z-X, Ji Q, Wible JR, Yuan C-X (2003) An Early Cretaceous tribosphenic mammal and metatherian

evolution. Science 302:1934–1940Mackness BS, Whitehead PW, McNamara GC (2000) New potassium–argon basalt date in relation to the

Pliocene Bluff Downs local fauna, northern Australia. Aust J Earth Sci 47:807–811Marshall LG (1976) New didelphine marsupials from the La-Venta Fauna Miocene of Colombia South America.

J Paleontol 50:402–418Marshall LG, de Muizon C (1988) The dawn of the age of mammals in South America. Natl Geograph Res 4:23–55Marshall LG, Case JA, Woodburne MO (1990) Phylogenetic relationships of the families of marsupials. In:

Genoways HH (ed) Current mammalogy, Vol 2. Plenum, New York, pp 433–506Marshall LG, Sempere T, Butler RF (1997) Chronostratigraphy of the mammal-bearing Paleocene of South

America. J South Am Earth Sci 10:49–70McKenna MC, Bell SK (1997) Classification of mammals above the species level. Columbia University Press,

New York

J Mammal Evol

Megirian D, Murray P, Schwartz L, Vonderborsch C (2004) Late Oligocene Kangaroo Well Local Fauna from theUlta Limestone (new name), and climate of the ‘Miocene oscillation’ across central Australia. Aust J EarthSci 51:701–741

Muizon C de, Cifelli RL (2001) A new basal “didelphoid” (Marsupialia, Mammalia) from the early Paleocene ofTiupampa (Bolivia). J Vertebr Paleontol 21:87–97

Muirhead J, Dawson L, Archer M (1997) Perameles bowensis, a new species of Perameles (Peramelemorphia:Marsupialia) from the Pliocene faunas of Bow and Wellington Caves, New South Wales. Proc Lin Soc NSW117:163–173

Müller J, Reisz RR (2005) Four well-constrained calibration points from the vertebrate fossil record formolecular clock estimates. BioEssays 27:1069–1075

Munemasa M, Nikaido M, Donnellan S, Austin CC, Okada N, Hasegawa M (2006) Phylogenetic analysis ofdiprotodontian marsupials based on complete mitochondrial genomes. Genes Genet Syst 81:181–191

Murphy WJ, Eizirik E, O’Brien SJ, Madsen O, Scally M, Douady CJ, Teeling E, Ryder OA, Stanhope MJ, deJong WW, Springer MS (2001) Resolution of the early placental mammal radiation using Bayesianphylogenetics. Science 294:2348–2351

Murray PF, Megirian D (1992) Continuity and contrast in middle and late Miocene vertebrate communities fromthe Northern Territory. Beagle 9:195–218

Myers TJ, Archer M (1997) Kuterintja ngama (Marsupialia, Ilariidae): A revised systematic analysis based onmaterial from the late Oligocene of Riversleigh, northwestern Queensland. Mem Qld Mus 41:379–392

Nilsson MA, Gullberg A, Spotorno AE, Arnason U, Janke A (2003) Radiation of extant marsupials after the K/Tboundary: evidence from complete mitochondrial genomes. J Mol Evol 57:S3–S12

Nilsson MA, Arnason U, Spencer PBS, Janke A (2004) Marsupial relationships and a timeline for marsupialradiation in south Gondwana. Gene 340:189–196

Osborne MJ, Christidis L (2001) Molecular phylogenetics of Australo-Papuan possums and gliders (FamilyPetauridae). Mol Phylogenet Evol 20:211–224

Osborne MJ, Christidis L, Norman JA (2002) Molecular phylogenetics of the Diprotodontia (kangaroos,wombats, koala, possums, and allies). Mol Phylogenet Evol 25:219–228

Penny D, Hasegawa M, Waddell PJ, Hendy MD (1999) Mammalian evolution: timing and implications fromusing the log determinant transform for proteins of differing amino acid composition. Syst Biol 48:76–93

Phillips MJ, Lin Y-H, Harrison GL, Penny D (2001) Mitochondrial genomes of a bandicoot and a brushtailpossum confirm the monophyly of australidelphian marsupials. Proc R Soc Lond B 268:1533–1538

Phillips MJ, McLenachan PA, Down C, Gibb GC, Penny D (2006) Combined mitochondrial and nuclear DNAsequences resolve the interrelations of the major Australasian marsupial radiations. Syst Biol 55:122–137

Pledge NS (1990) The upper fossil fauna of the Henschke fossil cave, Naracoorte, South Australia. Mem QueenslMus 28:247–262

Posada D, Crandall KA (1998) MODELTEST: testing the model of DNA substitution. Bioinformatics 14:817–818

Price GJ (2004) Fossil bandicoots (Marsupialia, Peramelidae) and environmental change during the Pleistoceneon the Darling Downs, southeastern Queensland, Australia. J Syst Palaeontol 2:347–356

Prideaux G (2004) Systematics and evolution of the sthenurine kangaroos. Univ Calif Pub Geol Sci 146:i–xvii,1-623, Berkeley, University of California Press

Rambaut A (1996) Se-Al: sequence alignment editor [computer editor]. Available at http://evolvezoooxacuk/Rambaut A, Drummond AJ (2003) Tracer version 12 [computer program]. Available http://evolvezoooxacuk

Accessed August 2006Reig OA (1952) Descripcion previa de nuevos ungulados y marsupiales fosiles del Plioceno y del eocuartario

Argentinos 1:119–129Reig OA (1958) Notas para una actualizacion del conocimiento de la fauna de la formacion Chapadmala II Amphibia,

Reptilia, Aves, Mammalia (Marsupialia: Didelphidae, Borhyaenidae). Acta Geol Lilloana, 2:255–283Reig OA, Kirsch JAW, Marshall LG (1987) Systematic relationships of the living and Neocenozoic American

“opossum-like” marsupials (Suborder Didelphimorphia) with comments on the classification of these and ofthe Cretaceous and Paleogene New World and European metatherians. In: Archer M (ed) Possums andopossums: studies in evolution. Surrey Beatty and Sons, Chipping Norton, New South Wales, pp 1–90

Reisz RR, Müller J (2004) Molecular timescales and the fossil record: a paleontological perspective. Trends Gen20:237–241

Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models.Bioinformatics 19:1572–1574

Rougier GW, Wible JR, Novacek MJ (1998) Implications of Deltatheridium specimens for early marsupialhistory. Nature 396:459–463

Sánchez-Villagra MR, Wible JR (2002) Patterns of evolutionary transformation in the petrosal bone and somebasicranial features in marsupial mammals, with special reference to didelphids. J Zool Syst Evol Res 40:26–45

J Mammal Evol

Sánchez-Villagra MR, Ladevèze S, Horovitz I, Argot C, Hooker JJ, Macrini TE, Martin T, Moore-Fay S,Muizon C de, Schmelzle T, Asher RJ (2007) Exceptionally preserved North American Paleogenemetatherians: adaptations and discovery of a major gap in the opossum fossil record. Biol Letters 3:318–322

Sanderson MJ (2002) Estimating absolute rates of molecular evolution and divergence times: a penalizedlikelihood approach. Mol Biol Evol 19:101–109

Schwartz L (2006) A new species of bandicoot from the Oligocene of Northern Australia and implications ofbandicoots for correlating Australian Tertiary mammal faunas. Palaeontol 49:991–998

Shimodaira H (2002) An approximately unbiased test of phylogenetic tree selection. Syst Biol 51:492–508Shimodaira H, Hasegawa M (1999) Multiple comparisons of log-likelihoods with applications to phylogenetic

inference. Mol Biol Evol 16:1114–1116Sigé B, Archer M, Godthelp H, Hand S, Crochet J-Y (1995) Peruvian–Australian Paleogene mammal

connection. Cavep’s 95, Fifth Conference on Australian Vertebrate Evolution, Palaeontology andSystematics, Canberra, Programme and Abstracts, 2

Sigé B, Sempere T, Butler RF, Marshall LG, Crochet J-Y (2004) Age and stratigraphic reassessment of the fossil-bearing Laguna Umayo red mudstone unit, SE Peru, from regional stratigraphy, fossil record, andpaleomagnetism. Geobios 37:771–794

Simpson GG (1974) Notes on Didelphidae (Mammalia, Marsupialia) from the Huayquerian (Pliocene) ofArgentina. Am Mus Novitates 2559:1–15

Smith AB, Pisani D, Mackenzie-Dodds JA, Stockley B, Webster BL, Littlewood TJ (2006) Testing the molecularclock: molecular and paleontological estimates of divergence times in the Echinoidea (Echinodermata). MolBiol Evol 23:1832–1851

Springer MS (1997) Molecular clocks and the timing of the placental and marsupial radiations in relation to theCretaceous–Tertiary boundary. J Mammal Evol 4:285–302

Springer MS, Kirsch JAW (1991) DNA hybridization, the compression effect, and the radiation of diprotodontianmarsupials. Syst Zool 40:131–151

Springer MS, Woodburne MO (1989) The distribution of some basicranial characters within the Marsupialia anda phylogeny of the Phalangeriformes. J Vertebr Paleontol 9:210–221

Springer MS, Kirsch JAW, Case JA (1997a) The chronicle of marsupial evolution. In: Givnish TJ, Sytsma KJ(eds) Molecular evolution and adaptive radiation. Cambridge University Press, New York, pp 129–161

Springer MS, Burk A, Kavanagh JR, Waddell VG, Stanhope MJ (1997b) The interphotoreceptor retinoid bindingprotein gene in therian mammals: implications for higher level relationships and evidence for loss of functionin the marsupial mole. Proc Natl Acad Sci USA 94:13754–13759

Springer MS, Westerman M, Kavanagh JR, Burk A, Woodburne MO, Kao DJ, Krajewski C (1998) The origin ofthe Australasian marsupial fauna and the phylogenetic affinities of the enigmatic monito del monte andmarsupial mole. Proc R Soc Lond B 265:2381–2386

Steiner C, Tilak M, Douzery JP, Catzeflis FM (2005) New DNA data from a transthyretin nuclear intron suggestan Oligocene to Miocene diversification of living South America opossums (Marsupialia: Didelphidae). MolPhylogenet Evol 35:363–379

Swofford DL (2002) PAUP* Phylogenetic analysis using parsimony (*and other methods), Version 4. SinauerAssociates, Sunderland, Massachusetts

Swofford DL, Olsen GJ, Waddell PJ, Hillis DM (1996) Phylogenetic inference. In: Hillis DM, Moritz C, MableBK (eds) Molecular systematics, 2nd edn. Sinauer, Sunderland, MA, pp 407–514

Szalay FS (1982) A new appraisal of marsupial phylogeny and classification. In: Archer M (ed) Carnivorousmarsupials. Royal Zoological Society of New South Wales, Sydney, pp 621–640

Szalay FS (1994) Evolutionary history of the marsupials and an analysis of osteological characters. CambridgeUniversity Press, New York

Szalay FS, Sargis EJ (2001) Model-based analysis of postcranial osteology of marsupials from the Palaeocene ofItaborai (Brazil) and the phylogenetics and biogeography of Metatheria. Geodiversitas 23:139–302

Tedford RH, Kemp NR (1998) Oligocene marsupials of the Geilston Bay local fauna, Tasmania. Am MusNovitates 3244:1–22

Teeling EC, Madsen O, Van Den Bussche RA, de Jong WW, Stanhope MJ, Springer MS (2002) Microbatparaphyly and the convergent evolution of a key innovation in Old World rhinolophoid microbats. Proc NatlAcad Sci USA 99:1431–1436

Teeling EC, Springer MS, Madsen O, Bates P, O’Brien SJ, Murphy WJ (2005) A molecular phylogeny for batsilluminates biogeography and the fossil record. Science 307:580–584

Temple-Smith PD (1987) Sperm structure and marsupial phylogeny. In: Archer M (ed) Possums and opossums:studies in evolution. Surrey Beatty and Sons, Chipping Norton, New South Wales, pp 171–193

Thorne JL, Kishino H (2002) Divergence time and evolutionary rate estimation with multilocus data. Syst Biol51:689–702

J Mammal Evol

Thorne JL, Kishino H, Painter IS (1998) Estimating the rate of evolution of the rate of molecular evolution. MolBiol Evol 15:1647–1657

Turnbull WD, Lundelius EL Jr, Archer M (2003) Dasyurids, perameloids, phalangeroids, and vombatoids fromthe early Pliocene Hamilton fauna, Victoria, Australia. Bull Am Mus Nat Hist 279:513–540

Warburton NM (2003) Functional morphology and evolution of marsupial moles (Marsupialia; Notoryctemor-phia). Thesis, The University of Western Australia

Westerman M, Springer MS, Dixon J, Krajewski C (1999) Molecular relationships of the extinct pig-footedbandicoot Chaeropus ecaudatus (Marsupialia: Perameloidea) using 12S rRNA sequences. J Mammal Evol6:271–288

Westerman M, Springer MS, Krajewski C (2001) Molecular relationships of the New Guinean bandicoot generaMicroperoryctes and Echymipera (Marsupialia:Peramelina). J Mammal Evol 8:93–105

White ME (1994) After the greening: the browning of Australia. Kangaroo, New South WalesWilson DE, Reeder DM (2005) Mammal species of the world: a taxonomic and geographic reference. The Johns

Hopkins University Press, BaltimoreWinge H (1941) The inter-relationships of the mammalian genera, Vol 1. Monotremata, Marsupialia, Insectivora,

Chiroptera, Edentata. CA Reitzels Forlag, CopenhagenWoodburne MO (1984) Families of marsupials: relationships, evolution, and biogeography. In: Broadhead TW

(ed) Mammals: notes for a short course. Univ Tenn Dept Geol Sci Stud Geol 8, pp 48–71Woodburne MO, Case JA (1996) Dispersal, vicariance, and the Late Cretaceous to early Tertiary land mammal

biogeography from South America to Australia. J Mammal Evol 3:121–161Woodburne MO, Tedford RH, Archer M, Turnbull WD, Plane MD, Lundelius EL Jr (1985) Biochronology of the

continental mammal record of Australia and New Guinea. Spec Publ S Aust Dept Mines Energy 5:347–364Woodburne MO, Tedford RH, Archer M (1987) New Miocene pseudocheirids (Pseudocheiridae: Marsupialia)

from South Australia. In: Archer M (ed) Possums and opossums: studies in evolution. Surrey Beatty andSons, Chipping Norton, New South Wales, pp 639–679

Woodburne MO, MacFadden BJ, Case JA, Springer MS, Pledge NS, Power JD, Woodburne JM, Springer KB(1993) Land mammal biostratigraphy and magnetostratigraphy of the Etadunna Formation (late Oligocene)of South Australia. J Vertebr Paleontol 13:483–515

Woodburne MO, Rich TH, Springer MS (2003) The evolution of tribospheny and the antiquity of mammalianclades. Mol Phylogenet Evol 28:360–385

Wroe S (1998) The geologically oldest dasyurid, from the Miocene of Riversleigh, north-west Queensland.Palaeontol 42:501–527

Wroe S (2003) Australian marsupial carnivores: recent advances in palaeontology. In: Jones M, Dickerman C,Archer M (eds) Predators with pouches: the biology of marsupial carnivores. CSIRO, Collingwood,Australia, pp 102–123

Wroe S, Mackness BS (2000) Additional material of Dasyurus dunmalli from the Pliocene Chinchilla LocalFauna of Queensland and its phylogenetic implications. Mem Queensl Mus 45:641–645

Wroe S, Ebach M, Ahyong S, Muizon C de, Muirhead J (2000) Cladistic analysis of dasyuromorphian(Marsupialia) phylogeny using cranial and dental characters. J Mammal 81:1008–1024

Yang Z, Rannala B (2006) Bayesian estimation of species divergence times under a molecular clock usingmultiple calibrations with soft bounds. Mol Biol Evol 23:212–226

J Mammal Evol